3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 5.14.0-284.18.1.el9 (RHEL9.2)
7 5.14.0-162.23.1.el9 (RHEL9.1)
8 4.18.0-477.15.1.el8 (RHEL8.8)
9 * Other server kernels known to build and work at some point (others may also work):
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 3.10.0-957.27.2.el7 (RHEL7.6)
12 3.10.0-1062.18.1.el7 (RHEL7.7)
13 3.10.0-1127.19.1.el7 (RHEL7.8)
14 3.10.0-1160.88.1.el7 (RHEL7.9)
15 4.18.0-80.11.2.el8 (RHEL8.0)
16 4.18.0-147.8.1.el8 (RHEL8.1)
17 4.18.0-193.28.1.el8 (RHEL8.2)
18 4.18.0-240.22.1.el8 (RHEL8.3)
19 4.18.0-305.25.1.el8 (RHEL8.4)
20 4.18.0-348.23.1.el8 (RHEL8.5)
21 4.18.0-372.32.1.el8 (RHEL8.6)
22 4.18.0-425.10.1.el8 (RHEL8.7)
23 4.4.120-92.70 (SLES12 SP2)
24 4.4.180-94.100 (SLES12 SP3)
25 4.4.0-131 (Ubuntu 16.04)
26 4.15.0-32 (Ubuntu 18.04)
27 5.4.0-48 (Ubuntu 20.04)
28 vanilla linux 5.4.0 (ZFS + ldiskfs)
29 vanilla linux 5.4.21 (ZFS + ldiskfs)
30 vanilla linux 5.4.136 (ZFS + ldiskfs)
31 5.10.0-60.94.0.118.oe2203 (openEuler 22.03 LTS)
32 5.10.0-136.32.0.108.oe2203sp1 (openEuler 22.03 LTS SP1)
33 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
34 * Client primary kernels built and tested during release cycle:
35 5.14.0-284.18.1.el9 (RHEL9.2)
36 5.14.0-162.23.1.el9 (RHEL9.1)
37 4.18.0-477.15.1.el8 (RHEL8.8)
38 5.4.0-37 (Ubuntu 20.04)
39 5.14.21-150400.24.46 (SLES15 SP4)
40 * Other clients known to build on these kernels at some point (others may also work):
41 3.10.0-862.14.4.el7 (RHEL7.5)
42 3.10.0-957.27.2.el7 (RHEL7.6)
43 4.14.0-49.13.1.el7a (RHEL7.5)
44 4.14.0-115.2.2.el7a (RHEL7.6)
45 3.10.0-1062.18.1.el7 (RHEL7.7)
46 3.10.0-1127.19.1.el7 (RHEL7.8)
47 3.10.0-1160.88.1.el7 (RHEL7.9)
48 4.18.0-80.11.2.el8 (RHEL8.0)
49 4.18.0-147.8.1.el8 (RHEL8.1)
50 4.18.0-193.28.1.el8 (RHEL8.2)
51 4.18.0-240.22.1.el8 (RHEL8.3)
52 4.18.0-305.25.1.el8 (RHEL8.4)
53 4.18.0-348.23.1.el8 (RHEL8.5)
54 4.18.0-372.32.1.el8 (RHEL8.6)
55 4.18.0-425.10.1.el8 (RHEL8.7)
56 5.14.0-70.30.1.el9 (RHEL9.0)
57 4.4.120-92.70 (SLES12 SP2)
58 4.4.180-94.100 (SLES12 SP3)
59 4.12.14-95.48 (SLES12 SP4)
60 4.12.14-122.133 (SLES12 SP5)
61 4.12.14-197.75 (SLES15 SP1)
62 5.3.18-24.96 (SLES15 SP2)
63 5.3.18-150300.59.93 (SLES15 SP3)
64 4.4.0-131 (Ubuntu 16.04)
65 4.15.0-48 (Ubuntu 18.04)
66 5.8.0-53 (Ubuntu 20.04.2 HWE)
67 5.11.0-31 (Ubuntu 20.04.3 HWE)
68 5.11.0 (vanilla kernel.org)
69 5.10.0 (openEuler 22.03 LTS)
70 * Recommended e2fsprogs version: 1.47.0-wc1 or newer
71 * Recommended ZFS version: 2.1.11
72 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
73 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
74 stack overflow. For more information, please refer to bugzilla 17630.
75 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
76 the re-exporting nodes to be mounted with "32bitapi" mount option
78 --------------------------------------------------------------------------------
83 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
84 for currently supported client and server kernel versions.
85 * Server primary kernels built and tested during release cycle:
86 4.18.0-348.2.1.el8 (RHEL8.5)
87 * Other server kernels known to build and work at some point (others may also work):
88 3.10.0-862.14.4.el7 (RHEL7.5)
89 3.10.0-957.27.2.el7 (RHEL7.6)
90 3.10.0-1062.18.1.el7 (RHEL7.7)
91 3.10.0-1127.19.1.el7 (RHEL7.8)
92 3.10.0-1160.49.1.el7 (RHEL7.9)
93 4.18.0-80.11.2.el8 (RHEL8.0)
94 4.18.0-147.8.1.el8 (RHEL8.1)
95 4.18.0-193.28.1.el8 (RHEL8.2)
96 4.18.0-240.22.1.el8 (RHEL8.3)
97 4.18.0-305.25.1.el8 (RHEL8.4)
98 4.4.120-92.70 (SLES12 SP2)
99 4.4.180-94.100 (SLES12 SP3)
100 4.4.0-131 (Ubuntu 16.04)
101 4.15.0-32 (Ubuntu 18.04)
102 5.4.0-48 (Ubuntu 20.04)
103 vanilla linux 5.4.0 (ZFS + ldiskfs)
104 vanilla linux 5.4.21 (ZFS + ldiskfs)
105 vanilla linux 5.4.136 (ZFS + ldiskfs)
106 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
107 * Client primary kernels built and tested during release cycle:
108 4.18.0-348.2.1.el8 (RHEL8.5)
109 5.4.0-37 (Ubuntu 20.04)
110 5.3.18-59.27 (SLES15 SP3)
111 * Other clients known to build on these kernels at some point (others may also work):
112 3.10.0-862.14.4.el7 (RHEL7.5)
113 3.10.0-957.27.2.el7 (RHEL7.6)
114 4.14.0-49.13.1.el7a (RHEL7.5)
115 4.14.0-115.2.2.el7a (RHEL7.6)
116 3.10.0-1062.18.1.el7 (RHEL7.7)
117 3.10.0-1127.19.1.el7 (RHEL7.8)
118 3.10.0-1160.49.1.el7 (RHEL7.9)
119 4.18.0-80.11.2.el8 (RHEL8.0)
120 4.18.0-147.8.1.el8 (RHEL8.1)
121 4.18.0-193.28.1.el8 (RHEL8.2)
122 4.18.0-240.22.1.el8 (RHEL8.3)
123 4.18.0-305.25.1.el8 (RHEL8.4)
124 4.4.120-92.70 (SLES12 SP2)
125 4.4.180-94.100 (SLES12 SP3)
126 4.12.14-95.48 (SLES12 SP4)
127 4.12.14-122.91 (SLES12 SP5)
128 4.12.14-197.75 (SLES15 SP1)
129 5.3.18-24.96 (SLES15 SP2)
130 4.4.0-131 (Ubuntu 16.04)
131 4.15.0-48 (Ubuntu 18.04)
132 5.8.0-53 (Ubuntu 20.04.2 HWE)
133 5.11.0-31 (Ubuntu 20.04.3 HWE)
134 5.11.0 (vanilla kernel.org)
135 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
136 * Recommended ZFS version: 2.0.7
137 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
138 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
139 stack overflow. For more information, please refer to bugzilla 17630.
140 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
141 the re-exporting nodes to be mounted with "32bitapi" mount option
143 --------------------------------------------------------------------------------
147 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
148 for currently supported client and server kernel versions.
149 * Server primary kernels built and tested during release cycle:
150 3.10.0-1062.18.1.el7 (RHEL7.7)
151 3.10.0-1127.19.1.el7 (RHEL7.8)
152 3.10.0-1160.6.1.el7 (RHEL7.9)
153 * Other server kernels known to build and work at some point (others may also work):
154 3.10.0-862.14.4.el7 (RHEL7.5)
155 3.10.0-957.27.2.el7 (RHEL7.6)
156 4.18.0-80.11.2.el8 (RHEL8.0)
157 4.18.0-147.8.1.el8 (RHEL8.1)
158 4.18.0-193.28.1.el8 (RHEL8.2)
159 4.18.0-240.1.1.el8 (RHEL8.3)
160 4.4.120-92.70 (SLES12 SP2)
161 4.4.180-94.100 (SLES12 SP3)
162 4.4.0-131 (Ubuntu 16.04)
163 4.15.0-32 (Ubuntu 18.04)
164 5.4.0-48 (Ubuntu 20.04)
165 vanilla linux 5.4.0 (ZFS + ldiskfs)
166 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
167 * Client primary kernels built and tested during release cycle:
168 3.10.0-1062.18.1.el7 (RHEL7.7)
169 3.10.0-1127.19.1.el7 (RHEL7.8)
170 3.10.0-1160.6.1.el7 (RHEL7.9)
171 4.12.14-95.48 (SLES12 SP4)
172 4.12.14-122.20 (SLES12 SP5)
173 4.12.14-197.75 (SLES15 SP1)
174 5.3.18-24.24 (SLES15 SP2)
175 4.15.0-48 (Ubuntu 18.04)
176 * Other clients known to build on these kernels at some point (others may also work):
177 3.10.0-862.14.4.el7 (RHEL7.5)
178 3.10.0-957.27.2.el7 (RHEL7.6)
179 4.14.0-49.13.1.el7a (RHEL7.5)
180 4.14.0-115.2.2.el7a (RHEL7.6)
181 4.18.0-80.11.2.el8 (RHEL8.0)
182 4.18.0-147.8.1.el8 (RHEL8.1)
183 4.18.0-193.28.1.el8 (RHEL8.2)
184 4.18.0-240.1.1.el8 (RHEL8.3)
185 4.4.120-92.70 (SLES12 SP2)
186 4.4.180-94.100 (SLES12 SP3)
187 4.4.0-131 (Ubuntu 16.04)
188 5.4.0-37 (Ubuntu 20.04)
189 5.4.0 (vanilla kernel.org)
190 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
191 * Recommended ZFS version: 2.0.0
192 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
193 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
194 stack overflow. For more information, please refer to bugzilla 17630.
195 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
196 the re-exporting nodes to be mounted with "32bitapi" mount option
198 --------------------------------------------------------------------------------
202 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
203 for currently supported client and server kernel versions.
204 * Server primary kernels built and tested during release cycle:
205 3.10.0-1062.1.1.el7 (RHEL7.7)
206 * Other server kernels known to build and work at some point (others may also work):
207 3.10.0-862.14.4.el7 (RHEL7.5)
208 3.10.0-957.27.2.el7 (RHEL7.6)
209 4.18.0-80.7.1.el8 (RHEL8.0)
210 4.4.120-92.70 (SLES12 SP2)
211 4.4.180-94.100 (SLES12 SP3)
212 4.4.0-131 (Ubuntu 16.04)
213 4.15.0-32 (Ubuntu 18.04)
214 vanilla linux 4.6.7 (ZFS only)
215 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
216 * Client primary unpatched kernels built and tested during release cycle:
217 3.10.0-1062.1.1.el7 (RHEL7.7)
218 4.12.14-95.29 (SLES12 SP4)
219 4.15.0-48 (Ubuntu 18.04)
220 * Other clients known to build on unpatched kernels at some point (others may also work):
221 3.10.0-862.14.4.el7 (RHEL7.5)
222 3.10.0-957.27.2.el7 (RHEL7.6)
223 4.14.0-49.13.1.el7a (RHEL7.5)
224 4.14.0-115.2.2.el7a (RHEL7.6)
225 4.18.0-80.7.1.el8 (RHEL8.0)
226 4.4.120-92.70 (SLES12 SP2)
227 4.4.180-94.100 (SLES12 SP3)
228 4.4.0-131 (Ubuntu 16.04)
229 vanilla linux 4.15.0 (ZFS only)
230 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
231 * Recommended ZFS / SPL version: 0.7.13
232 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
233 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
234 stack overflow. For more information, please refer to bugzilla 17630.
235 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
236 the re-exporting nodes to be mounted with "32bitapi" mount option
237 * Note for Ubuntu 14.04.5:
238 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
239 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
241 --------------------------------------------------------------------------------
245 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
246 for currently supported client and server kernel versions.
247 * Server known to build on the following kernels (others may also work):
248 2.6.32-696.18.7.el6 (RHEL6.9)
249 3.10.0-862.14.4.el7 (RHEL7.5)
250 3.10.0-957.el7 (RHEL7.6)
251 4.4.120-92.70 (SLES12 SP2)
252 4.4.162-94.69 (SLES12 SP3)
253 4.4.0-131 (Ubuntu 16.04)
254 4.15.0-32 (Ubuntu 18.04)
255 vanilla linux 4.6.7 (ZFS only)
256 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
257 * Client known to build on unpatched kernels (others may also work):
258 2.6.32-696.18.7.el6 (RHEL6.9)
259 2.6.32-754.6.3.el6 (RHEL6.10)
260 3.10.0-862.14.4.el7 (RHEL7.5)
261 3.10.0-957.el7 (RHEL7.6)
262 4.14.0-49.13.1.el7a (RHEL7.5)
263 4.14.0-115.2.2.el7a (RHEL7.6)
264 4.4.120-92.70 (SLES12 SP2)
265 4.4.162-94.69 (SLES12 SP3)
266 4.4.0-131 (Ubuntu 16.04)
267 4.15.0-32 (Ubuntu 18.04)
268 vanilla linux 4.15.0 (ZFS only)
269 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
270 * Recommended ZFS / SPL version: 0.7.9
271 * Tested with ZFS / SPL version: 0.7.9
272 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
273 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
274 stack overflow. For more information, please refer to bugzilla 17630.
275 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
276 the re-exporting nodes to be mounted with "32bitapi" mount option
277 * Note for Ubuntu 14.04.5:
278 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
279 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
281 --------------------------------------------------------------------------------
283 2018-04-03 Intel Corporation
285 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
286 for currently supported client and server kernel versions.
287 * Server known to build on patched kernels:
288 2.6.32-431.29.2.el6 (RHEL6.5)
289 2.6.32-504.30.3.el6 (RHEL6.6)
290 2.6.32-573.26.1.el6 (RHEL6.7)
291 2.6.32-642.15.1.el6 (RHEL6.8)
292 2.6.32-696.18.7.el6 (RHEL6.9)
293 3.10.0-693.21.1.el7 (RHEL7.4)
294 3.0.101-0.47.71 (SLES11 SP3)
295 3.0.101-107 (SLES11 SP4)
296 3.12.74-60.64.40 (SLES12 SP1)
297 4.4.114-92.64 (SLES12 SP2)
298 4.4.114-94.11 (SLES12 SP3)
299 3.13.0-101 (Ubuntu 14.04, ZFS only)
300 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
301 4.4.0-116 (Ubuntu 16.04)
302 vanilla linux 4.6.7 (ZFS only)
303 * Client known to build on unpatched kernels:
304 2.6.32-431.29.2.el6 (RHEL6.5)
305 2.6.32-504.30.3.el6 (RHEL6.6)
306 2.6.32-573.26.1.el6 (RHEL6.7)
307 2.6.32-642.15.1.el6 (RHEL6.8)
308 2.6.32-696.18.7.el6 (RHEL6.9)
309 3.10.0-693.21.1.el7 (RHEL7.4)
310 3.0.101-0.47.71 (SLES11 SP3)
311 3.0.101-107 (SLES11 SP4)
312 3.12.74-60.64.40 (SLES12 SP1)
313 4.4.114-92.64 (SLES12 SP2)
314 4.4.114-94.11 (SLES12 SP3)
315 3.13.0-101 (Ubuntu 14.04)
316 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
317 4.4.0-116 (Ubuntu 16.04)
319 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
320 * Recommended ZFS / SPL version: 0.7.6
321 * Tested with ZFS / SPL version: 0.7.6
322 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
323 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
324 stack overflow. For more information, please refer to bugzilla 17630.
325 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
326 the re-exporting nodes to be mounted with "32bitapi" mount option
327 * Note for Ubuntu 14.04.5:
328 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
329 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
331 --------------------------------------------------------------------------------
333 2017-07-05 Intel Corporation
335 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
336 for currently supported client and server kernel versions.
337 * Server known to build on patched kernels:
338 2.6.32-431.29.2.el6 (RHEL6.5)
339 2.6.32-504.30.3.el6 (RHEL6.6)
340 2.6.32-573.26.1.el6 (RHEL6.7)
341 2.6.32-642.15.1.el6 (RHEL6.8)
342 2.6.32-696.3.1.el6 (RHEL6.9)
343 3.10.0-514.21.1.el7 (RHEL7.3)
344 3.0.101-0.47.71 (SLES11 SP3)
345 3.0.101-100 (SLES11 SP4)
346 3.12.74-60.64.40 (SLES12 SP1)
347 4.4.59-92.17 (SLES12 SP2)
348 vanilla linux 4.6.7 (ZFS only)
349 * Client known to build on unpatched kernels:
350 2.6.32-431.29.2.el6 (RHEL6.5)
351 2.6.32-504.30.3.el6 (RHEL6.6)
352 2.6.32-573.26.1.el6 (RHEL6.7)
353 2.6.32-642.15.1.el6 (RHEL6.8)
354 2.6.32-696.3.1.el6 (RHEL6.9)
355 3.10.0-514.21.1.el7 (RHEL7.3)
356 3.0.101-0.47.71 (SLES11 SP3)
357 3.0.101-100 (SLES11 SP4)
358 3.12.74-60.64.40 (SLES12 SP1)
359 4.4.59-92.17 (SLES12 SP2)
361 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
362 * Recommended ZFS / SPL version: 0.6.5.9
363 * Tested with ZFS / SPL version: 0.6.5.9
364 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
365 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
366 stack overflow. For more information, please refer to bugzilla 17630.
367 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
368 the re-exporting nodes to be mounted with "32bitapi" mount option
370 --------------------------------------------------------------------------------
372 12-07-2016 Intel Corporation
374 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
375 for currently supported client and server kernel versions.
376 * Server known to build on patched kernels:
377 2.6.32-431.29.2.el6 (RHEL6.5)
378 2.6.32-504.30.3.el6 (RHEL6.6)
379 2.6.32-573.26.1.el6 (RHEL6.7)
380 2.6.32-642.11.1.el6 (RHEL6.8)
381 3.10.0-514.el7 (RHEL7.3)
382 3.0.101-0.47.71 (SLES11 SP3)
383 3.0.101-80 (SLES11 SP4)
384 3.12.62-60.64.8 (SLES12 SP1)
385 vanilla linux 4.5.7 (ZFS only)
386 * Client known to build on unpatched kernels:
387 2.6.32-431.29.2.el6 (RHEL6.5)
388 2.6.32-504.30.3.el6 (RHEL6.6)
389 2.6.32-573.26.1.el6 (RHEL6.7)
390 2.6.32-642.11.1.el6 (RHEL6.8)
391 3.10.0-514.el7 (RHEL7.3)
392 3.0.101-0.47.71 (SLES11 SP3)
393 3.0.101-80 (SLES11 SP4)
394 3.12.62-60.64.8 (SLES12 SP1)
396 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
397 * Recommended ZFS / SPL version: 0.6.5.8
398 * Tested with ZFS / SPL version: 0.6.5.8
399 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
400 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
401 stack overflow. For more information, please refer to bugzilla 17630.
402 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
403 the re-exporting nodes to be mounted with "32bitapi" mount option
405 --------------------------------------------------------------------------------
407 02-29-2016 Intel Corporation
409 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
410 for currently supported client and server kernel versions.
411 * Server known to build on patched kernels:
412 2.6.32-431.29.2.el6 (RHEL6.5)
413 2.6.32-504.30.3.el6 (RHEL6.6)
414 2.6.32-573.12.1.el6 (RHEL6.7)
415 3.10.0-327.3.1.el7 (RHEL7.2)
416 3.0.101-0.47.71 (SLES11 SP3)
417 3.0.101-68 (SLES11 SP4)
418 vanilla linux 4.2.1 (ZFS only)
419 * Client known to build on unpatched kernels:
420 2.6.32-431.29.2.el6 (RHEL6.5)
421 2.6.32-504.30.3.el6 (RHEL6.6)
422 2.6.32-573.12.1.el6 (RHEL6.7)
423 3.10.0-327.3.1.el7 (RHEL7.2)
424 3.0.101-0.47.71 (SLES11 SP3)
425 3.0.101-68 (SLES11 SP4)
428 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
429 * Recommended ZFS / SPL version: 0.6.4.2
430 * Tested with ZFS / SPL version: 0.6.4.2
431 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
432 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
433 stack overflow. For more information, please refer to bugzilla 17630.
434 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
435 the re-exporting nodes to be mounted with "32bitapi" mount option
437 --------------------------------------------------------------------------------
439 03-10-2015 Intel Corporation
441 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
442 for currently supported client and server kernel versions.
443 * Server known to build on patched kernels:
444 2.6.32-431.29.2.el6 (RHEL6.5)
445 2.6.32-504.8.1.el6 (RHEL6.6)
446 3.0.101-0.46 (SLES11 SP3)
447 * Client known to build on unpatched kernels:
448 2.6.32-431.29.2.el6 (RHEL6.5)
449 2.6.32-504.8.1.el6 (RHEL6.6)
450 3.10.0-123.20.1.el7 (RHEL7)
451 3.0.101-0.46 (SLES11 SP3)
452 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
453 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
454 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
455 stack overflow. For more information, please refer to bugzilla 17630.
456 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
457 the re-exporting nodes to be mounted with "32bitapi" mount option
459 Severity : enhancement
461 Description: control OST-index in IDIF via ROCOMPAT flag.
462 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
463 last_rcvd file. For new formatted OST device, it will be auto set;
464 for the case of upgrading from old OST device, you can enable it
465 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
466 enabled, for new created OST-object, its IDIF-in-LMA will contain
467 the OST-index; for the existing OST-object, the OSD will convert
468 old format IDIF as new format IDIF with OST-index stored in the
469 LMA EA when accessing such OST-object or via OI scrub. Once such
470 flag is enabled, it cannot be reverted back, so the system cannot
471 be downgraded to the orignal incompatible version.
473 --------------------------------------------------------------------------------
475 07-30-2014 Intel Corporation
477 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
478 for currently supported client and server kernel versions.
479 * Server known to build on patched kernels:
480 2.6.32-431.20.3.el6 (RHEL6.5)
481 3.0.101-0.31 (SLES11 SP3)
482 * Client known to build on unpatched kernels:
483 2.6.32-431.20.3.el6 (RHEL6.5)
484 3.0.101-0.31 (SLES11 SP3)
487 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
488 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
489 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
490 stack overflow. For more information, please refer to bugzilla 17630.
491 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
492 the re-exporting nodes to be mounted with "32bitapi" mount option
494 --------------------------------------------------------------------------------
496 10-31-2013 Intel Corporation
498 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
499 * Server builds for kernels:
500 2.6.32-358.18.1.el6 (RHEL6.4)
501 * Client builds for unpatched kernels:
502 2.6.32-358.18.1.el6 (RHEL6.4)
503 3.0.92-0.8 (SLES11 SP3)
505 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
506 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
507 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
508 stack overflow. For more information, please refer to bugzilla 17630.
509 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
510 the re-exporting nodes to be mounted with "32bitapi" mount option
512 --------------------------------------------------------------------------------
514 05-30-2013 Intel Corporation
516 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
517 * Server builds for kernels:
518 2.6.32-358.6.2.el6 (RHEL6.4)
519 * Client builds for unpatched kernels:
520 2.6.32-358.6.2.el6 (RHEL6.4)
521 3.0.74-0.6.6 (SLES11 SP2)
523 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
524 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
525 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
526 stack overflow. For more information, please refer to bugzilla 17630.
527 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
528 the re-exporting nodes to be mounted with "32bitapi" mount option
530 --------------------------------------------------------------------------------
532 9-30-2012 Whamcloud, Inc.
534 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
535 * Server support for kernels:
536 2.6.32-279.5.1.el6 (RHEL6)
537 * Client support for unpatched kernels:
538 2.6.32-279.5.1.el6 (RHEL6)
540 2.6.18-238.19.1.el5 (RHEL5)
541 2.6.16.60-0.69.1 (SLES10)
542 * Recommended e2fsprogs version: 1.41.90.wc4
543 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
544 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
545 stack overflow. For more information, please refer to bugzilla 17630.
546 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
547 the re-exporting nodes to be mounted with "32bitapi" mount option
549 Severity : enhancement
551 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
553 Severity : enhancement
555 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
557 Severity : enhancement
559 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
561 Severity : enhancement
563 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
565 Severity : enhancement
567 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
569 --------------------------------------------------------------------------------
571 04-30-2012 Whamcloud, Inc.
573 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
574 * Server support for kernels:
575 2.6.32-220.4.2.el6 (RHEL6)
576 * Client support for unpatched kernels:
577 2.6.32-220.4.2.el6 (RHEL6)
579 2.6.18-238.19.1.el5 (RHEL5)
580 2.6.16.60-0.69.1 (SLES10)
581 * Recommended e2fsprogs version: 1.41.90.wc4
582 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
583 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
584 stack overflow. For more information, please refer to bugzilla 17630.
585 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
586 the re-exporting nodes to be mounted with "32bitapi" mount option
588 Severity : enhancement
590 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
592 Severity : enhancement
594 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
596 Severity : enhancement
598 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
600 Severity : enhancement
602 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
604 --------------------------------------------------------------------------------
606 09-30-2011 Whamcloud, Inc.
608 * Server support for kernels:
609 2.6.18-238.19.1.el5 (RHEL5)
610 2.6.32-131.6.1.el6 (RHEL6)
611 * Client support for unpatched kernels:
612 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
613 2.6.18-238.19.1.el5 (RHEL5)
614 2.6.32-131.6.1.el6 (RHEL6)
615 2.6.29.4-167.fc11 (FC11)
616 2.6.16.60-0.69.1 (SLES10)
618 * Recommended e2fsprogs version: 1.41.90.wc3
619 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
620 Lustre filesystem with 4K stack may cause a stack overflow. For more
621 information, please refer to bugzilla 17630.
622 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
623 mounted with "32bitapi" mount option
625 Severity : enhancement
626 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
628 Severity : enhancement
629 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
631 Severity : enhancement
632 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
635 Frequency : only with SLES10
637 Description: Use OFED "KMP" provided by Novell
638 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
639 Lustre is now built against this package. That means you need to
640 install the ofed-kmp package from Novell for the patchless client.
641 Note that the ofed-kmp that Novell ships may not exactly match the
642 kernel version but should still be compatible.
644 Severity : enhancement
646 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
647 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
649 --------------------------------------------------------------------------------
651 2010-07-15 Oracle, Inc.
653 * Server support for kernels:
654 2.6.18-164.11.1.el5 (RHEL 5)
655 2.6.18-164.11.1.0.1.el5 (OEL 5)
656 * Client support for unpatched kernels:
657 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
658 2.6.18-164.11.1.el5 (RHEL 5),
659 2.6.18-164.11.1.0.1.el5 (OEL 5)
660 2.6.16.60-0.42.8 (SLES 10),
662 2.6.29.4-167.fc11 (FC11)
663 * Recommended e2fsprogs version: 1.41.10-sun2
665 Severity : enhancement
667 Description: Distribute lfs_migrate script with lustre
668 Details : The lfs_migrate mentioned in the manual is now included in the
669 lustre tree in lustre/scripts and will distributed in lustre RPMs.
671 Severity : enhancement
673 Description: Failover nid list added to import proc file.
675 Severity : enhancement
677 Description: "som_preview" mount option required to use SOM
678 Details : The clients will need to use som_preview mount option
682 Description: Lustre does not do 1MB IOs to HW RAID
683 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
685 Severity : enhancement
687 Description: lfs quota output is not very convenient for awk/sed-parsing
688 Details : Some positions in lfs quota output table could be empty or
689 non-empty which made it hard to parse it with scripts, now
690 a dash is put instead of space where there is not supposed
695 Description: Update to OFED 1.5.1
696 Details : OFED 1.5.1 will be used in environments that don't use
697 vendor supplied OFED.
701 Description: Remove lnet. prefix from lctl params display
702 Details : Also renamed lnet.memused to lnet_memused to
707 Description: Use conf_param -d to remove permanant settings.
711 Description: Add list_param -R
712 Details : lctl list_param now supports recursive mode.
715 Bugzilla : 21829, 22312
716 Description: Fix llobdstat.
717 Details : 1. Now updating the write_bytes counter
718 2. Port script fixes from 1.8.
722 Description: decrease the usage of memory on clients.
723 Details : 1. On clients, recycle dentries and inodes unused.
724 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
729 Description: llobdstat fix and enhancement
730 Details : add a counter to set a limit to how many samples will be returned
731 fix a wildcard in the path to limit to obdfilter stats only
735 Description: lfs check only allowed for root
736 Details : Code cleanups and added a regresion test of non-root use of
741 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
742 Details : Fix hr->hr_index overload.
746 Description: Fix FIEMAP functionality.
750 Description: FC11 patchless client support added
754 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
755 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
759 Description: New acc-sm script and various test-framework enhancments.
763 Description: Previously landed patch was reverted.
767 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
771 Description: Add ldiskfs maxdirsize mount option.
775 Description: Fix mkfs.lustre support for 16TB LUNs.
779 Description: Print mmp_check_interval
780 Details : print mmp_check_interval and make it possible to abort
781 mount operation in case it takes too long.
785 Description: Add cascading_rw to lustre mpi tests.
789 Description: Store SOM data in EA.
790 Details : Also several SOM related cleanups.
794 Description: Add refcount to md_open_data
798 Description: Use refcount on lov_request_set to prevent memory corruption.
802 Description: Statahead: use parent dir's i_mutex for synchronization.
806 Description: loadgen improvements.
810 Description: Enable peer health detection in ksocklnd and ko2iblnd.
814 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
818 Description: Fix race on obd_devcie:md_stats
822 Description: Fix SLES 11 client builds.
823 Details : Resolved uninitalized pointer gcc warnings.
827 Description: Allow non-root access for "lfs check".
828 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
832 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
834 Severity : enhancement
836 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
837 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
841 Description: Sanity test_27p fails.
842 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
843 Sanity tests 27 were cleaned up and improved.
847 Description: Unify req format on client/servers
848 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
849 buffers and swabbers approach..
854 Description: drop join file stuff.
856 Severity : enhancement
858 Description: Add 2.6.27 and SLES11 (patchless client) support
862 Description: Fix in ptlrpc_expire_one_request() to print the signed time
865 Severity : enhancement
867 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
870 Severity : enhancement
872 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
874 Severity : enhancement
876 Description: correctly handle big reply message.
877 Details : send LNet event if reply is bigger than buffer and adjust this buffer
882 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
884 Severity : enhancement
886 Description: Add support for OEL5.
890 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
892 Severity : enhancement
894 Description: "weak-modules" support
895 Details : Implement "weak-modules" support which enables kernel modules
896 to be used with any kernel that implements the same kABI. In
897 order to achieve this modules are now installed in
898 /lib/modules/$(uname -r)/updates/kernel on all distributions.
902 Description: client could not reconnect to OST because of an active request.
903 Details : The client now retries to reconnect to the same server, if a
904 connect request failed with EBUSY or -EAGAIN. Test case 35b
905 conf-sanity.sh has been created.
909 Description: Don't allow make backward step on assignin osc next id.
910 Details : race between allocation next id and ll_sync thread can be cause
911 of set wrong osc next id and can be kill valid ost objects.
915 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
919 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
923 Description: Error handling in osc_statfs_interpret() has been improved.
924 Details : Check in osc_statfs_interpret() for EBADR.
928 Description: Do not send statfs() requests to OSTs disabled by administrator.
929 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
931 Severity : enhancement
933 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
935 Severity : enhancement
937 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
939 Severity : enhancement
941 Description: Enhancement for lfs(1) command to use numeric uid/gid.
945 Description: Improved cleanup procedure for conf-sanity test_46a.
949 Description: Do not start bulk transfer, if client tried to reconnect.
950 Details : Do not start bulk transfer, if client tried to reconnect,
951 but failed due to active rpcs. Check for exp_abort_active_req
952 has been implemented.
956 Description: Prevent inconsistences between linux and lustre mount structures.
957 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
958 Make the sleep interruptible.
960 Severity : enhancement
962 Description: provide server to client comms path
963 Details : server to client communications path via new LDLM_SET_INFO rpc
967 Description: truncate starts GFP_FS allocation under transaction causing deadlock
968 Details : ldiskfs_truncate calls grab_cache_page which may start page
969 allocation under an open transaction. This may lead to
970 calling prune_icache with consequent lustre reentrance.
974 Description: send correctly lsm on open replay
975 Details : MDS is trust to LSM size on replay open, but client can set wrong size
980 Description: enable client interface failover
981 Details : When a child reconnects from another NID, properly update export
982 nid hash position and ldlm reverse import.
984 Severity : enhancement
986 Description: Communicate OST degraded/readonly state via statfs to MDS
987 Details : Flags in the statfs returned from OSTs indicate whether the
988 OST is in a degraded RAID state, or if the filesystem has
989 turned read-only after a filesystem error is detected.
994 Description: don't panic if EPROTO was hit when reading symlink
995 Details : correctly handling request reference in error cases.
997 Severity : enhancement
999 Description: Add LustreNetLink, a kernel-userspace communcation path.
1001 Severity : enhancement
1003 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1008 Description: access to llog context before init.
1009 Details : move handling CATALOGS file at osc layer and forbid access to llog
1010 context before init.
1013 Frequency : rare, only when client is writing to OST on local node
1015 Description: Avoid deadlock for local client writes
1016 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1017 memory freeing context. This allows OST threads to set the
1018 PF_MEMALLOC flag on task structures in order to allocate memory
1019 from reserved pools and complete IO.
1020 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1021 so that the OST threads generate memory pressure and allow
1022 inactive pages to be reclaimed.
1024 Severity : enhancement
1026 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1031 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1032 Details : move ldlm namespace creation in setup phase to avoid grab
1033 _lprocfs_lock with cli_sem held.
1037 Description: Temporarily disable grant shrink.
1038 Details : Disable the feature for debugging.
1041 Frequency : only when formatting test filesystems
1043 Description: Unable to run several mkfs.lustre on loop devices at the same time
1044 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1045 formatting. The solution is to proper handle the error.
1047 Severity : enhancement
1049 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1051 Severity : enhancement
1053 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1058 Description: resolve race between obd_disconnect and class_disconnect_exports
1059 Details : if obd_disconnect will be called to already disconnected export he
1060 forget release one reference and osc module can't unloaded.
1062 Severity : enhancement
1064 Description: Allow tuning service thread via /proc
1065 Details : For each service a new
1066 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1067 created that can be used to set min/max thread counts, and get the
1068 current number of running threads.
1073 Descriptoin: don't return error if have partially created objects for file.
1074 Details : lov_update_create_set uses set->set_success as index for created
1075 objects, so if some requests will be failed, they will have hole
1076 at end of array and qos_shrink_lsm for allocate correct lsm.
1078 Severity : enhancement
1080 Description: Update OFED support to 1.4.1
1085 Description: Reduce small size read RPC
1086 Details : Set read-ahead limit for every file and only do read-ahead when
1087 available read-ahead pages are bigger than 1M to avoid small size
1090 Severity : enhancement
1092 Description: move AT tunable parameters for more consistent usage
1093 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1095 Severity : enhancement
1097 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1098 Details : allow skip disconnected ost for send statfs request and hide error
1104 Description: fix racy locking of mballoc block bitmaps causing BUG
1105 Details : The locking of the mballoc buddy bitmap and the in-memory
1106 block bitmap was using two different spin locks in some
1107 cases. This made it possible to incorrectly access the
1108 mballoc bitmap while another process was modifying it,
1109 causing a sanity assertion to fail. While no on-disk corruption
1110 was reported, there was some risk of this happening.
1113 Frequency : rare, on llog test 6
1115 Descriptoin: don't allow connect to already connected import
1116 Details : allowing connect to already connected import is hide connecting problem.
1121 Descriptoin: allow kill process which wait statahead result
1122 Details : for some reasons 'ls' can stick in waiting result from statahead,
1123 in this case need way for kill this process.
1125 Severity : enhancement
1127 Description: Add state history info file, enhance import info file
1128 Details : Track import connection state changes in a new osc/mdc proc file;
1129 add overview-type data to the osc/mdc import proc file.
1131 Severity : enhancement
1133 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1135 Severity : enhancement
1137 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1142 Descriptoin: OST grant too much space to client even there are not enough space.
1143 Details : Client will shrink its grant cache to OST if there are no write
1144 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1145 this grant cache if there are already not enough avaible space
1146 (left_space < total_clients * 32M).
1149 Frequency : start MDS on uncleanly shutdowned MDS device
1151 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1152 Details : stay in waiting mds<>ost recovery finished produce random bugs
1153 due race between two ll_sync thread for one lov target. send
1154 ACTIVATE event only if connect realy finished and import have
1158 Frequency : rare, connect and disconnect target at same time
1160 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1161 Details : don't call obd_disconnect under lov_lock. this long time
1162 operation and can block ptlrpcd which answer to connect request.
1167 Descriptoin: don't lose wakeup for imp_recovery_waitq
1168 Details : recover_import_no_retry or invalidate_import and import_close can
1169 both sleep on imp_recovery_waitq, but we was send only one wakeup
1173 Frequency : always with long access acl
1175 Descriptoin: mds can't pack reply with long acl.
1176 Details : mds don't control size of acl but they limited by reint/getattr
1179 Severity : enhancement
1181 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1183 Severity : enhancement
1185 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1188 Frequency : start MDS on uncleanly shutdowned MDS device
1190 Descriptoin: aborting recovery hang on MDS
1191 Details : don't throttle destroy RPCs for the MDT.
1194 Frequency : on remount
1196 Description: external journal device not working after the remount
1197 Details : clear dev_rdonly flag for external journal devices in
1203 Description: shutdown vs evict race
1204 Details : client_disconnect_export vs connect request race.
1205 if client will evicted at this time - we start invalidate
1206 thread without referece to import and import can be freed
1210 Frequency : rare, need acl's on inode.
1212 Description: client can't handle ost additional correctly
1213 Details : if ost was added after client connected to mds client can have
1214 hit lnet_try_match_md ... to big messages to wide striped files.
1215 in this case need teach client to handle config events about add
1216 lov target and update client max ea size at that event.
1218 Severity : enhancement
1220 Description: Changelogs
1221 Details : Changelogs are a lightweight mechanism to track filesystem
1222 metadata and namespace changes. The changelog is recorded
1223 permanently on the MDTs, and is periodically "consumed" / purged
1224 when records are no longer needed.
1226 Severity : enhancement
1228 Description: compact fld format with extents
1229 Details : Store range of seq rather than every seq in FLD. Seq
1230 controller update FLD rather than clients. In Case of CMD, mdt0
1231 has FLD, all other metadata server act as non persistent proxy
1232 for FLD queries and cache fld entries in fld cache.
1237 Description: don't skip ost target if they assigned to file
1238 Details : Drop slow OSCs if we can, but not for requested start idx.
1239 This means "if OSC is slow and it is not the requested
1240 start OST, then it can be skipped, otherwise skip it only
1241 if it is inactive/recovering/out-of-space.
1245 Description: more cleanup in mds_lov
1246 Details : not send LOV EA under replay, we can't know about they size at this
1247 time. Don't allow client connect to mds before any ost connected,
1248 for avoid problems with LOV EA size and returning EIO to client.
1250 Severity : enhancement
1252 Description: Interoperability at server side (Disk interoperability)
1254 Severity : enhancement
1256 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1258 Severity : enhancement
1260 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1262 Severity : enhancement
1264 Description: New client IO stack (CLIO).
1266 Severity : enhancement
1268 Description: Commit on sharing. Eliminate inter-client dependencies between
1269 uncommitted transactions by doing transaction commits.
1270 Thereby clients may recovery independently.
1273 Frequency : Create a symlink file with a very long name
1275 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1276 Details : If there is no extra space in the request for early cancels,
1277 ldlm_req_handles_avail() returns 0 instead of a negative value.
1279 Severity : enhancement
1281 Description: Add /proc entry for import status
1282 Details : The mdc, osc, and mgc import directories now have
1283 an import directory that contains useful import data for debugging
1284 connection problems.
1286 Severity : enhancement
1288 Description: Re-disable certain /proc logging
1289 Details : Enable and disable client's offset_stats, extents_stats and
1290 extents_stats_per_process stats logging on the fly.
1293 Frequency : Only on FC kernels 2.6.22+
1295 Description: oops in statahead
1296 Details : Do not drop reference count for the dentry from VFS when lookup,
1297 VFS will do that by itself.
1299 Severity : enhancement
1301 Description: Generic /proc file permissions
1302 Details : Set /Proc file permissions in a more generic way to enable non-
1303 root users operate on some /proc files.
1307 Description: Hitting mdc_commit_close() ASSERTION
1308 Details : Properly handle request reference release in
1309 ll_release_openhandle().
1313 Description: quota recovery deadlock during mds failover
1314 Details : This patch includes att18982, att18236, att18237 in bz14840.
1316 1. fix osts hang when mds does failover with quotaon
1317 2. prevent watchdog storm when osts threads wait for the
1322 Frequency : only patchless client
1323 Description: add workaround for race between add/remove dentry from hash
1325 Severity : enhancement
1327 Description: Allow OST glimpses to return PW locks
1331 Description: LBUG when llog conf file is full
1332 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1336 Description: Prevent import from entering FULL state when server in recovery
1340 Description: service mount cannot take device name with ":"
1341 Details : Only when device name contains ":/" will mount treat it as
1347 Description: replace ptlrpcd with the statahead thread to interpret the async
1348 statahead RPC callback
1352 Frequency : on recovery
1353 Description: I/O failures after umount during fail back
1354 Details : if client reconnected to restarted server we need join to recovery
1355 instead of find server handler is changed and process self
1356 eviction with cancel all locks.
1358 Severity : enhancement
1360 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1362 Severity : enhancement
1364 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1366 Severity : enhancement
1368 Description: Upcall on Lustre log has been dumped
1369 Details : Allow for a user mode script to be called once a Lustre log has
1370 been dumped. It passes the filename of the dumped log to the
1371 script, the location of the script can be specified via
1372 /proc/sys/lnet/debug_log_upcall.
1377 Description: avoid idr_remove called for id which is not allocated.
1378 Details : Move assigment s_dev for clustered nfs to end of initialization,
1379 to avoid problem with error handling.
1384 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1385 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1386 into CONN_USED_HASH and this prodice warning when put connection
1387 again in unused hash.
1390 Severity : enhancement
1392 Description: File striping can now be set to use an arbitrary pool of OSTs.
1394 Severity : enhancement
1396 Description: Export bytes_read/bytes_write count on OSC/OST.
1400 Description: Early reply size mismatch, MGC loses connection
1401 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1402 the connect flags are properly negotiated.
1407 Description: quotas are not honored with O_DIRECT
1408 Details : all writes with the flag O_DIRECT will use grants which leads to
1409 this problem. Now using OBD_BRW_SYNC to guard this.
1413 Description: add quota statistics
1414 Details : 1. sort out quota proc entries and proc code.
1415 2. add quota statistics
1417 Severity : enhancement
1419 Description: enable quota support for HEAD.
1423 Description: Properly propagate oinfo flags from lov to osc for statfs
1424 Details : restore missing copy oi_flags to lov requests.
1426 Severity : enhancement
1428 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1429 : l_getgroups(8), lst(8), routerstat(8)
1430 Details : included man pages for llobdstat(8), llstat(8),
1431 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1433 Severity : enhancement
1435 Description: configure's --enable-quota should check the
1436 : kernel .config for CONFIG_QUOTA
1437 Details : configure is terminated if --enable-quota is passed but
1438 : no quota support is in kernel
1442 Description: Remove portals compatibility
1443 Details : Remove portals compatibility, not interoperable with releases
1448 Description: Resolve device initialization race
1449 Details : Prevent proc handler from accessing devices added to the
1450 obd_devs array but yet be intialized.
1452 Severity : enhancement
1454 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1456 Severity : enhancement
1458 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1463 Description: Using wrong pointer in osc_brw_prep_request
1464 Details : Access to array[-1] can produce panic if kernel compiled with
1465 CONFIG_PAGE_ALLOC enabled
1469 Description: Client runs out of low memory
1470 Details : Consider only lowmem when counting initial number of llap pages
1474 Description: *optional* service tags registration
1475 Details : if the "service tags" package is installed on a Lustre node
1476 When the filesystem is mounted, a local-node service tag will
1477 be created. See http://inventory.sun.com/ for more information
1478 about the Service Tags asset management system.
1482 Description: Kernel BUG tries to release flock
1483 Details : Lustre does not destroy flock lock before last reference goes
1484 away. So always drop flock locks when client is evicted and
1485 perform unlock regardless of successfulness of speaking to MDS.
1489 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1493 Description: sanity test 65a fails if stripecount of -1 is set
1494 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1499 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1500 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1503 Severity : enhancement
1505 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1509 Description: do not process already freed flock
1510 Details : flock can possibly be freed by another thread before it reaches
1511 to ldlm_flock_completion_ast.
1515 Description: LBUG during stress test
1516 Details : Need properly lock accesses the flock deadlock detection list.
1520 Description: oops in page fault handler
1521 Details : kernel page fault handler can return two special 'pages' in error
1522 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1526 Description: timeout with invalidate import.
1527 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1528 should be handled by ptlrpcd. This produce long age waiting and
1529 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1531 Severity : enhancement
1533 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1537 Description: enable MGS and MDT services start separately
1538 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1539 a co-located MGS without starting the MGS, which is a complement
1540 to 'nosvc' mount option.
1544 Frequency : after recovery
1545 Description: precreate to many object's after del orphan.
1546 Details : del orphan st in oscc last_id == next_id and this triger growing
1547 count of precreated objects. Set flag LOW to skip increase count
1548 of precreated objects.
1552 Frequency : rare, on clear nid stats
1553 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1554 Details : when clean nid stats sometimes try destroy live entry,
1555 and this produce panic in free.
1559 Description: Stack overflow during MDS log replay
1560 ease stack pressure by using a thread dealing llog_process.
1564 Description: wait until IO finished before start new when do lock cancel.
1565 Details : VM protocol want old IO finished before start new, in this case
1566 need wait until PG_writeback is cleared until check dirty flag and
1567 call writepages in lock cancel callback.
1569 Severity : enhancement
1571 Description: using special macro for print time and cleanup in includes.
1575 Description: mds_mfd_close() ASSERTION(rc == 0)
1576 Details : In mds_mfd_close(), we need protect inode's writecount change
1577 within its orphan write semaphore to prevent possible races.
1581 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1585 Frequency : rare, on shutdown ost
1586 Description: don't hit live lock with umount ost.
1587 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1588 use shrink_dcache_sb instead.
1592 Description: don't panic with use echo client
1593 Details : echo client pass NULL as client nid pointer and this produce null
1594 pointer dereference.
1598 Description: fix build on ppc32
1599 Details : compile code with -m64 flag produce wrong object file for ppc32.
1603 Description: add message levels for liblustreapi
1607 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1608 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1609 will result in an -EAGAIN and in the exit call path we call
1610 alloc_rr() which will with increasing aggressiveness attempt to
1611 aquire precreated objects on the minimum number of required OSCs.
1615 Description: Use old size assignment to avoid deadlock
1616 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1617 the scheduling while holding a spinlock. We do not need locking
1618 for size in ll_update_inode() because size is only updated from
1619 the MDS for directories or files without objects, so there is no
1620 other place to do the update, and concurrent access to such inodes
1621 are protected by the inode lock.
1625 Description: resolve "_IOWR redefined" build error on SLES10
1629 Description: dump the memory debugging after all modules are unloaded to
1630 suppress false negative in conf_sanity test 39
1632 Severity : enhancement
1634 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1639 Description: fix SLES kernel versioning
1640 Details : the kernel version for our SLES 10 kernel did not include a "-"
1641 before the "smp" at the end. while this was not a problem in
1642 general, it did mean that software trying to use the kernel
1643 version to try to detect a vendor specific kernel would fail.
1644 this was most evident by the OFED build scripts.
1648 Description: Don't update lov_desc members until making sure they are valid
1649 Details : When updating lov_desc members via proc fs, need fix their
1650 validities before doing the real update.
1654 Description: don't put request into delay list while invalidate in flight.
1655 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1656 import in flight. this produce timeout for invalidate and sometimes
1657 can cause stale data.
1661 Frequency : on ppc only
1662 Description: not convert ost objects for directory because it's not exist.
1663 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1667 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1669 Severity : enhancement
1671 Description: Update to support 2.6.22.14 vanilla kernel.
1675 Frequency : rare, on recovery
1676 Description: read procfs can produce deadlock in some situation
1677 Details : Holding lprocfs lock which send rpc can produce block for destroy
1678 obd objects and this also block reconnect with -EALREADY. This
1679 isn't fix all lprocfs bugs - but make it rare.
1681 Severity : enhancement
1683 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1686 Frequency : frequent on X2 node
1688 Description: mdc_set_open_replay_data LBUG
1689 Details : Set replay data for requests that are eligible for replay.
1693 Description: lustre_mgs: operation 101 on unconnected MGS
1694 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1695 MGC, and late on MGC cannot successfully connect to MGS and a lot
1696 of the error messages complaining that MGS is not connected.
1700 Frequency : on network error
1701 Description: panic with double free request if network error
1702 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1703 but it's true only for synchronus enqueue, for async enqueue
1704 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1707 Severity : enhancement
1709 Description: client-side metadata stat-ahead during readdir(directory readahead)
1710 Details : perform client-side metadata stat-ahead when the client detects
1711 readdir and sequential stat of dir entries therein
1714 Frequency : on start mds
1716 Description: Implement get_info(last_id) in obdfilter.
1719 Frequency : occasional
1721 Description: Correctly check stale fid, not start epoch if ost not support SOM
1722 Details : open with flag O_CREATE need set old fid in op_fid3 because
1723 op_fid2 was overwritten with new generated fid, but mds can answer
1724 with one of these two fids and both is not stale. Setattr
1725 incorrectly started an epoch and assume will be called
1726 done_writing, but without SOM done_writing ever being called.
1729 Frequency : rare, depends on device drivers and load
1731 Description: MDS or OSS nodes crash due to stack overflow
1732 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1733 In some cases, in conjunction with device drivers that use a lot
1734 of stack the MDS (or possibly OSS) service threads could overflow
1735 the stack. One change which was identified to consume additional
1736 stack has been reworked to avoid the extra stack usage.
1739 Frequency : occasional
1741 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1742 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1743 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1744 oscc_has_objects and see OSC has no objests and call
1745 oscc_internal_create to resend create request.
1747 Severity : enhancement
1749 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1751 Severity : enhancement
1753 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1756 Frequency : very rare
1758 Description: Fix replay if there is an un-replied request and open
1759 Details : In some cases, older replay request will revert the
1760 mcd->mcd_last_xid on MDS which is used to record the client's
1761 latest sent request.
1763 Severity : enhancement
1765 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1767 Severity : enhancement
1769 Description: Add rhel5 support to HEAD.
1771 Serverity : enhancement
1773 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1778 Description: Don't allow skipping OSTs if index has been specified.
1779 Details : Don't allow skipping OSTs if index has been specified, make
1780 locking in internal create lots better.
1784 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1785 Details : access to bitfield in structure is always rounded to long
1786 and this produce problem with not atomic change any bit.
1790 Description: Lustre make rpms failed.
1791 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1792 builds Lustre rpms from tarball.
1795 Frequency : rare on shutdown ost
1797 Description: If llog cancel was not send before clean_exports phase, this can
1798 produce deadlock in llog code.
1799 Details : If llog thread has last reference to obd and call class_import_put
1800 this produce deadlock because llog_cleanup_commit_master wait when
1801 last llog_commit_thread exited, but this never success because was
1802 called from llog_commit_thread.
1806 Description: allow userland application know is lost one of stripes.
1807 Details : fill lvb_blocks with error code on ost and return it to
1808 application if error flag found.
1812 Description: NULL lov_tgts causing MDS oops
1813 Details : more safe checks for NULL lov_tgts for avoid oops.
1815 Severity : enhancement
1817 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1821 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1822 Details : If an OST is down the MDS will hang indefinitely in
1823 obd_statfs_rqset() waiting for the statfs data. While for
1824 MDS QOS usage of statfs, it should not stuck in waiting.
1826 Severity : enhancement
1828 Description: remote_acl support
1829 Details : Support ACL-based permission check for remote user.
1830 Support setfacl/getfacl for remote user with the utils
1831 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1832 the system "{s,g}etfacl" utils.
1834 Severity : enhancement
1836 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1838 Severity : enhancement
1840 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1844 Description: make mgs_setparam() handle fsname containing dash
1845 Details : fsname containing a dash does not work with lctl conf_param
1847 Severity : enhancement
1849 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1851 Severity : enhancement
1853 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1857 Frequency : rare, at shutdown
1858 Description: access already free / zero obd_namespace.
1859 Details : if client_disconnect_export was called without force flag set,
1860 and exist connect request in flight, this can produce access to
1861 NULL pointer (or already free pointer) when connect_interpret
1862 store ocd flags in obd_namespace.
1866 Frequency : only at startup
1867 Description: not alloc memory with spinlock held.
1868 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1869 if any spinlock held.
1871 Severity : enhancement
1873 Description: make lustre randomly fail allocating memory
1874 Details : Make lustre randomly failed allocating memory for testing purpose.
1876 Severity : enhancement
1878 Description: lost problems with lov objid file
1879 Details : Fixes some scability and access to not inited memory problems
1880 in work with lov objdid file.
1885 Description: lfs find does not continue on file error
1886 Details : Continue other files processing when a file/dir is absent.
1890 Description: Inconsistent usage of lustre_pack_reply()
1891 Details : Standardize the usage of lustre_pack_reply() such that it
1892 always generate a CERROR on failure.
1895 Frequency : occasional
1897 Description: MDS hang or stay in waiting lock
1898 Details : If client receive lock with CBPENDING flag ldlm need send lock
1899 cancel as separate rpc, to avoid situation when cancel request
1900 can't processed due all i/o threads stay in wait lock.
1904 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1908 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1912 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1916 Description: Update to RHEL4 latest kernel.
1918 Severity : enhancement
1920 Description: Build SLES10 patchless client fails
1921 Details : The configure was broken by run ./configure with
1922 --with-linux-obj=.... argument for patchless client. When the
1923 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1924 can't search header adequately. Use absolute path such as
1925 -I($LINUX)/include instead.
1929 Description: interrupt oig_wait produce painc on resend.
1930 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1931 private set, and this produce situation when rq_ptlrpcd_data not
1932 copyed to new allocated request and triggered LBUG on assert
1933 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1934 ptlrpcd set if request is changed and can be safety changed to use
1939 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1940 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1941 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1942 these macros and replace them with their difinitions instead.
1946 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1947 Details : Update which_patch & target file for SLES10 latest kernel.
1950 Description: Remove client patches from SLES 10 kernel.
1951 Details : This causes SLES 10 clients to behave as patchless clients
1952 even on a Lustre-patched (server) kernel.
1954 Severity : enhancement
1956 Description: self-adjustable client's lru lists
1957 Details : use adaptive algorithm for managing client cached locks lru
1958 lists according to current server load, other client's work
1959 pattern, memory activities, etc. Both, server and client
1960 side namespaces provide number of proc tunables for controlling
1963 Severity : enhancement
1965 Description: light-weight GSS support
1966 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1967 and reduce performance overhead.
1969 Severity : enhancement
1971 Description: Linux keyring support
1972 Details : Support using service of Linux keyring for Lustre GSS internal
1973 context refresh/cache mechanism.
1977 Description: Fix errors in lfs documentation
1978 Details : Fixes man pages
1982 Description: don't use GFP_* in generic Lustre code.
1983 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1987 Description: obdclass is limited by single OBD_ALLOC(idarray)
1988 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1992 Description: warnings with build patchless client with vanila 2.6.19 and up
1993 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1994 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1998 Frequency : Only for SLES
1999 Description: sanity_quota fail test_1
2000 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2001 file, which makes TSTID[2] inunique.
2005 Frequency : Only for SLES9
2006 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2007 Details : Update target/ChangeLog/which_patch .
2011 Frequency : Only for test_10 in sanity-quota.sh
2012 Description: a bug in quota test code
2013 Details : checking a test flag in a wrong place
2017 Frequency : Only for test_13 in sanity-quota.sh
2018 Description: a bug in quota test script
2019 Details : shouldn't check overfull allocated quota
2023 Description: jbd statistics
2024 Details : Port older jbd statistics patch for sles10
2026 Severity : enhancement
2028 Description: Add printing inode info into message about error in writepage.
2033 Description: reply_lock_interpret crash due to race with it and lock cancel.
2034 Details : Do not replay locks that are being cancelled. Do not reference
2035 locks by their address during replay, just by their handle.
2039 Frequency : When flocks are used.
2040 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2041 Details : Flock locks might destroy just granted lock if it could be merged
2042 with another existing flock, this is done in completion handler,
2043 so teach ldlm_cli_enquque_fini that this is a valid case for
2049 Description: Oops in read and write path when failing to allocate lock.
2050 Details : Check if lock allocation failed and return error back.
2054 Description: lstripe command fails for valid OST index
2055 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2056 instead of lov->desc.ld_active_tgt_count.
2060 Description: sanity.sh failed test 103
2061 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2066 Description: df doesn't work properly if diskfs blocksize != 4K
2067 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2071 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2072 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2074 Severity : enhancement
2076 Description: Adaptive timeouts
2077 Details : RPC timeouts adapt to changing server load and network
2078 conditions to reduce resend attempts and improve recovery time.
2082 Description: llapi_file_create() does not allow some changes
2083 Details : add llapi_file_open() that allows specifying the mode and
2084 open flags, and also returns an open file handle.
2088 Description: merge and cleanup kernel patches.
2089 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2093 Description: Add journal checksum support.(Kernel part)
2094 Details : The journal checksum feature adds two new flags i.e
2095 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2096 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2097 indicates that the commit block contains the checksum for
2098 the blocks described by the descriptor blocks. Now commit
2099 record can be sent to disk without waiting for descriptor
2100 blocks to be written to disk. This behavior is controlled
2101 using JBD2_FEATURE_ASYNC_COMMIT flag.
2105 Description: OSS needs mutliple precreate threads
2106 Details : Add ability to start more than one create thread per OSS.
2110 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2111 Details : execute lfs setstripe on client
2115 Description: mds_obd_create error creating tmp object
2116 Details : When the user sets quota on root, llog will be affected and can't
2117 create files and write files.
2120 Frequency : Always on ia64 patchless client, and possibly others.
2122 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2123 Details : This allows the patchless client to be loaded on architectures
2124 without this export.
2128 Description: RedHat Update kernel for RHEL5
2129 Details : Add the kernel config file for RHEL5.
2133 Description: RedHat Update kernel for RHEL5
2134 Details : Modify the kernel config file more closer RHEL5.
2138 Description: Build failure against Centos5 (RHEL5)
2139 Details : Define PAGE_SIZE when it isn't present.
2143 Description: test_7 on sanity-quota.sh can't be used on separate machines
2144 Details : add facet to handle it
2148 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2149 Details : with NFS, the anon dentry's parent was set to itself in
2150 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2151 corresponding dentry other than use rec->ur_name.
2153 Severity : enhancement
2155 Description: lfs setstripe enhancement
2156 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2160 Description: enable data checksumming by default
2161 Details : enable checksum by default, allow --disable-checksum
2162 configure option and "-o nochecksum" mount option. Checksums
2163 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2167 Description: lustre support for RHEL5
2168 Details : Add support for RHEL5.
2172 Description: Client eviction due to failover config
2173 Details : after a connection loss, the lustre client should attempt to
2174 reconnect to the last active server first before trying the
2175 other potential connections.
2179 Description: when mds and osts use different quota unit(32bit and 64bit),
2180 quota will be released repeatly.
2181 Details : void sending multiple quota reqs to mds, which will keep the
2182 status between the reqs.
2186 Description: osts not allocated evenly to files
2187 Details : change the condition to increase offset_idx
2190 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2192 Description: Data corruption for OSTs that are formatted larger than 2TB
2194 Details : When generating the bio request for lustre file writes the
2195 sector number would overflow a temporary variable before being
2196 used for the IO. The data reads correctly from Lustre (which
2197 will overflow in a similar manner) but other file data or
2198 filesystem metadata may be corrupted in some cases.
2202 Description: Tune the kernel for good SCSI performance.
2203 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2204 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2209 Description: rewrite ext2-derived code in obdclass/uuid.c
2210 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2211 from scratch preserving functionality.
2215 Description: "lfs find -obd UUID" prints directories
2216 Details : "lfs find -obd UUID" will return all directory names instead
2217 of just file names. It is incorrect because the directories
2218 do not reside on the OSTs.
2222 Description: use i_size_read and i_size_write in 2.6 port
2223 Details : replace inode->i_size access with i_size_read/write()
2226 Frequency : when removing large files
2228 Description: scheduling issue during removal of large Lustre files
2229 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2230 It causes scheduling issues when removing large files (17TB in the
2234 Frequency : only with liblustre clients on XT3
2236 Description: evictions taking too long
2237 Details : allow llrd to evict clients directly on OSTs
2240 Frequency : only on ppc
2242 Description: /proc/fs/lustre/devices broken on ppc
2243 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2244 We should make sure the type of 'index' is loff_t and then cast
2245 explicitly as needed below. Do not assign an explicitly cast
2249 Frequency : only for rhel5
2251 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2252 Details : Modify the target file & which_kernel.
2254 Severity : enhancement
2256 Description: omit set fsid for export NFS
2257 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2258 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2262 Frequency : Always, for kernels after 2.6.16
2263 Description: Fix warning idr_remove called for id=.. which is not allocated.
2264 Details : Last kernels save old s_dev before kill super and not allow
2265 to restore from callback - restore it before call kill_anon_super.
2269 Description: buffer overruns could theoretically occur
2270 Details : llapi_semantic_traverse() modifies the "path" argument by
2271 appending values to the end of the origin string, and a
2272 overrun may occur. Adding buffer overrun check in liblustreapi.
2276 Description: Fix error on 'ls .' at the top of the Lustre mount.
2277 Details : Don't revalidate dentry if it is a root dentry.
2281 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2282 Details : Modify vm-tunables-rhel4.patch.
2286 Description: Kernel config for 2.6.18-vanilla.
2287 Details : Modify targets/2.6-vanilla.target.in.
2288 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2289 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2290 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2291 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2295 Description: improve handling recoverable errors
2296 Details : if request processig with error which can be recoverable on server
2297 request should be resend, otherwise page released from cache and
2302 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2303 Details : Modify target file & which_patch.
2304 A flaw was found in the IA32 system call emulation provided
2305 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2306 value could be stored in the %RAX register, which could trigger an
2307 out-of-bounds system call table access. An untrusted local user
2308 could exploit this flaw to run code in the kernel
2309 (ie a root privilege escalation). (CVE-2007-4573).
2313 Description: O_DIRECT bypasses client statistics.
2314 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2315 read_ahead_stats were updated. copied the stats section from
2316 osc_send_oap_rpc() into async_internal().
2320 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2324 Description: change order of libsysio includes
2325 Details : '#include sysio.h' should always come before '#include xtio.h'
2329 Description: handle "serious error: objid * already exists" more gracefully
2330 Details : If LAST_ID value on disk is smaller than the objects existing in
2331 the O/0/d* directories, it indicates disk corruption and causes an
2332 LBUG(). If the object is 0-length, then we should use the existing
2333 object. This will help to avoid a full fsck in most cases.
2335 Severity : enhancement
2337 Description: adapt the lustre_config script to support the upgrade case
2338 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2339 server targets to 1.6 in parallel.
2343 Description: To avoid grant space > avaible space when the disk is almost
2344 full. Without this patch you might see the error "grant XXXX >
2345 available" or some LBUG about grant, when the disk is almost
2347 Details : In filter_check_grant, for non_grant cache write, we should
2348 check the left space by if (*left > ungranted + bytes), instead
2349 of (*left > ungranted), because only we are sure the left space
2350 is enough for another "bytes", then the ungrant space should be
2351 increase. In client, we should update cl_avail_grant only there
2352 is OBD_MD_FLGRANT in the reply.
2356 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2357 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2358 described in CVE-2007-4573. This problem could allow untrusted
2359 local users to gain root access.
2362 Frequency : when using O_DIRECT and quotas
2364 Description: Incorrect file ownership on O_DIRECT output files
2365 Details : block usage reported by 'lfs quota' does not take into account
2366 files that have been written with O_DIRECT.
2371 Description: touch file failed when fs is not full
2372 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2373 otherwise we can get ENOSP while fs is not full.
2377 Description: parallel lock callbacks
2378 Details : Instead of sending blocking and completion callbacks as separated
2379 requests, adding them to a set and sending in parallel.
2382 Frequency : only for Cray XT3
2383 Bugzilla : 12829/13455
2384 Description: Changing primary group doesn't change the group lustre assigns to
2386 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2387 overridden with the primary group provided by the group upcall,
2388 whereas the supplied fsgid can be trusted if it is in the list of
2389 supplementary groups returned by the group upcall.
2391 Severity : enhancement
2393 Description: Allow masking D_WARNING, D_ERROR messages from console
2394 Details : Console messages can now be disabled via lnet.printk.
2398 Description: User code with malformed file open parameter crashes client node
2399 Details : Before packing join_file req, all the related reference should be
2400 checked carefully in case some malformed flags cause fake
2401 join_file req on client.
2405 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2406 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2407 mode is equal to the granted mode. However, it should also check
2408 that LDLM_FL_CP_REQD is not set.
2412 Description: Heavy nfs access might result in deadlocks
2413 Details : After ELC code landed, it is now improper to enqueue any mds
2414 locks under och_sem, because enqueue might want to decide to
2415 cancel open locks for same inode we are holding och_sem for.
2419 Description: Client eviction while running blogbench
2420 Details : A lot of unlink operations with concurrent I/O can lead to a
2421 deadlock causing evictions. To address the problem, the number of
2422 oustanding OST_DESTROY requests is now throttled to
2423 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2424 callbacks are processed in priority.
2428 Description: enable ACLs on MDS by default
2429 Details : ACLs must be enabled on MDS by default.
2432 Frequency : PPC/PPC64 only
2434 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2435 Details : fix duplicated definitions between asm-ppc64/types.h and
2436 lustre_types.h on PPC.
2439 Frequency : PPC/PPC64 only
2441 Description: asm-ppc/segment.h does not exist
2442 Details : fix compile issue on PPC.
2446 Description: better handle error messages in extents code
2449 Frequency : RHEL4 only
2451 Description: mkfs is very slow on IA64/RHEL4
2452 Details : A performance regression has been discovered in the MPT Fusion
2453 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2454 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2455 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2457 Severity : enhancement
2459 Description: SNMP support enhancement
2460 Details : Adding total number of sampled request for an MDS node in snmp
2463 Severity : enhancement
2465 Description: Optimize ldlm waiting list processing for PR extent locks
2466 Details : When processing waiting list for read extent lock and meeting read
2467 lock that is same or wider to it that is not contended, skip
2468 processing rest of the list and immediatelly return current
2469 status of conflictness, since we are guaranteed there are no
2470 conflicting locks in the rest of the list.
2474 Description: Time out and refuse to reconnect
2475 Details : When the failover node is the primary node, it is possible
2476 to have two identical connections in imp_conn_list. We must
2477 compare not conn's pointers but NIDs, otherwise we can defeat
2478 connection throttling.
2482 Description: port llog fixes from b1_6 into HEAD
2483 Details : Port llog reference couting and some llog cleanups from b1_6
2484 (bug 10800) into HEAD, for protect from panic and access to already
2485 free llog structures.
2489 Description: Detect stride IO mode in read-ahead
2490 Details : When a client does stride read, read-ahead should detect that and
2491 read-ahead pages according to the detected stride pattern.
2495 Description: data checksumming impacts single node performance
2496 Details : add support for several checksum algorithm. Currently, only CRC32
2497 and Adler-32 are supported. The checksum type can be changed on
2498 the fly via /proc/fs/lustre/osc/*/checksum_type.
2502 Description: use adler32 for page checksums
2503 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2508 Description: build for x2 fails
2509 Details : fix compile issue on Cray systems.
2513 Description: Properly match for duplicate locks
2514 Details : Due to different lock order from skiplists code, we need to
2515 traverse entire list for now
2518 Frequency : only on PPC/SLES10
2520 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2521 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2522 be defined. Add a hack in mkfs_lustre.c to work around this compile
2527 Description: LASSERT on MDS when client holding flock lock dies
2528 Details : ldlm pool logic depends on number of granted locks equal to
2529 number of released locks which is not true for flock locks, so
2530 just exclude such locks from consideration.
2534 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2535 Details : Use fsfilt_sync() for both the whole filesystem sync and
2536 individual file sync to eliminate dangerous inode locking
2537 with I_LOCK that can lead to a deadlock.
2541 Description: performance in 1.6.3
2542 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2543 sure that 1MB requests are not fragmented by the block layer.
2545 Severity : enhancement
2547 Description: organize the server-side client stats on per-nid basis
2548 Details : Change the structure of stats under obdfilter and mds to
2556 The "uuid"s file would list the uuids of _active_ exports.
2557 And the clear entry is to clear all stats and stale nids.
2559 Severity : enhancement
2561 Description: eliminate client locks in face of contention
2562 Details : file contention detection and lockless i/o implementation
2563 for contended files.
2567 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2570 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2572 Description: MDS may lose file striping (and hence file data) in some cases
2573 Details : If there are additional extended attributes stored on the MDS,
2574 in particular ACLs, SELinux, or user attributes (if user_xattr
2575 is specified for the client mount options) then there is a risk
2576 of attribute loss. Additionally, the Lustre file striping
2577 needs to be larger than default (e.g. striped over all OSTs),
2578 and an additional attribute must be stored initially in the
2579 inode and then increase in size enough to be moved to the
2580 external attribute block (e.g. ACL growing in size) for file
2585 Description: skiplist implementation simplification
2586 Details : skiplists are used to group compatible locks on granted list
2587 that was implemented as tracking first and last lock of each
2588 lock group the patch changes that to using doubly linked lists
2592 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2593 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2594 because LASSERT on both the data supplied by a client, and the
2595 data on disk is dangerous and incorrect.
2597 Severity : enhancement
2599 Description: Slow truncate/writes to huge files at high offsets.
2600 Details : Directly associate cached pages to lock that protect those pages,
2601 this allows us to quickly find what pages to write and remove
2602 once lock callback is received.
2606 Description: more ldlm soft lockups
2607 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2608 starve other threads from the resource lock for a long time in
2609 case of long waiting queue, so change the debug level from
2610 D_OTHER to the less frequently used D_INFO.
2612 Severity : enhancement
2614 Description: add -gid, -group, -uid, -user options to lfs find
2618 Description: Hung threads in invalidate_inode_pages2_range
2619 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2620 once one is completed. As a result, some RPCs are stuck in the
2621 queue and are never sent.
2625 Description: filter threads hungs on waiting journal commit
2626 Details : Cleanup filter group llog code, then only filter group llog will
2627 be only created in the MDS/OST syncing process.
2631 Description: Procfs and llog threads access destoryed import sometimes.
2632 Details : Sync the import destoryed process with procfs and llog threads by
2633 the import refcount and semaphore.
2635 Severity : enhancement
2637 Description: openlock cache of b1_6 port to HEAD
2642 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2643 Details : If insertion of an extent fails, then discard the inode
2644 preallocation and free data blocks else it can lead to duplicate
2649 Description: don't always update ctime in ext3_xattr_set_handle()
2650 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2651 In some cases the ctime should not be updated, for example for
2652 2.0->1.8 compatibility it is necessary to delete an xattr and it
2653 should not update the ctime.
2657 Bugzilla : 15713/16362
2658 Description: Assertion in iopen_connect_dentry in 1.6.3
2659 Details : looking up an inode via iopen with the wrong generation number can
2660 populate the dcache with a disconneced dentry while the inode
2661 number is in the process of being reallocated. This causes an
2662 assertion failure in iopen since the inode's dentry list contains
2663 both a connected and disconnected dentry.
2667 Description: assertion failure in ldlm_handle2lock()
2668 Details : fix a race between class_handle_unhash() and class_handle2object()
2669 introduced in lustre 1.6.5 by bug 13622.
2674 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2675 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2680 Description: oops in sd_iostats_seq_show()
2681 Details : unloading/reloading the scsi low level driver triggers a kernel
2682 bug when trying to access the sd iostat file.
2687 Description: Kernel panics during QLogic driver reload
2688 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2689 patch, causing memory corruption.
2694 Description: journal_dev option does not work in b1_6
2695 Details : pass mount option during pre-mount.
2697 Severity : enhancement
2699 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2700 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2701 extent information of a file. It can be used to map logical blocks
2702 in a file to physical blocks in the block device.
2706 Description: LDLM soft lockups - improvement
2707 Details : It is be possible to send the lock handle along with each read
2708 or write request because the client is already doing a lock match
2709 itself so there isn't any reason the OST should have to re-do that
2713 Frequency : only on Cray X2
2715 Description: X2 build failures
2716 Details : fix build failures on Cray X2.
2720 Description: xid & resent requests
2721 Details : Initialize RPC XID from clock at startup (randomly if clock is
2724 Severity : enhancement
2726 Description: Add lustre_start utility to start or stop multiple Lustre servers
2731 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2732 Details : In case of memory pressure, list_del() can be called twice on
2733 req->rq_history_list, causing a kernel oops.
2737 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2738 Details : fix stack overflow in the distributed lock manager by defering
2739 export eviction after a failed AST to the elt thread instead of
2740 handling it in the dlm interpret routine.
2744 Description: Convert some comments to new format.
2745 Details : Update documenting comments to match doxygen conventions.
2749 Description: Grammar fixes.
2750 Details : A couple of trivial spelling fixes.
2754 Description: OSD_COUNTERS-mandatory
2755 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2756 to simplify the code.
2760 Description: simplify cmm_device freeing logic.
2761 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2765 Description: Add lockdep support to dt_object_operations locking interface.
2766 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2767 parameter indicating lock ordering. Update mdd code to use new
2772 Description: Introduce failloc constants for lockless IO tests.
2773 Details : Add two new failloc constants to test lockless IO. Only one of
2774 them in implemented---another is checked in yet to be landed
2779 Description: Add lockdep support for inode mutex.
2780 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2781 in the situations where Lustre has to lock more than one inode
2786 Description: Add optional invariants checking support.
2787 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2788 configure switch. This macro is to be used for consistency and
2789 sanity checks that are too expensive to be left in `production'
2794 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2795 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2800 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2801 Details : Use spin_lock_nested() in (the only) situation where more than
2802 one ldlm_lock is locked simultaneously. Also, fix possible
2803 dead-lock in ldlm_lock_change_resource() by enforcing particular
2808 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2809 Details : Instead of specifying each ldlm_lock call-back through separate
2810 parameter, wrap them into struct ldlm_callback_suite.
2814 Description: Kill join_lru obd method and its callers.
2815 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2816 in memory---a requirement implemented through ->o_join_lru() obd
2817 method in HEAD. Remove this method and its users.
2821 Description: Add asynchronous ldlm ENQUEUE completion handler.
2822 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2823 case that stock ldlm_completion_ast() cannot handle as it waits
2824 until lock is granted. Introduce new ldlm_completion_ast_async()
2825 for this. Also comment ldlm_completion_ast().
2829 Description: ldlm_error <-> errno conversion.
2830 Details : Add functions to map (rather arbitrary) between LDLM error codes
2831 and standard errno values. CLIO needs this to prevent LDLM specific
2832 constants from escaping ldlm and osc.
2836 Description: Kill unused ldlm_handle2lock_ns() function.
2837 Details : Kill unused ldlm_handle2lock_ns() function.
2841 Description: Add lu_ref support to ldlm_lock
2842 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2843 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2844 and ldlm_resource. LDLM interface has to be changed, because code that
2845 releases a reference on a lock, has to "know" what reference this is.
2846 In the most frequent case
2848 lock = ldlm_handle2lock(handle);
2850 LDLM_LOCK_PUT(lock);
2852 no changes are required. When any other reference (received _not_ from
2853 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2854 instead of LDLM_LOCK_PUT().
2856 Arguably, changes are pervasive, and interface requires some discipline
2857 for proper use. On the other hand, it was very instrumental in finding
2858 a few leaked lock references.
2862 Description: Add ldlm_lock_addref_try().
2863 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2864 attempts to addref a lock that might be being canceled
2869 Description: Add ldlm_weigh_callback().
2870 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2871 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2872 hard-coded `number of pages' logic.
2876 Description: Add lockdep annotations to llog code.
2877 Details : Use appropriately tagged _nested() locking calls in the places
2878 where llog takes more than one ->lgh_lock lock.
2882 Description: Add loi_kms_set().
2883 Details : Wrap kms updates into a helper function.
2887 Description: Constify instances of struct lsm_operations.
2888 Details : Constify instances of struct lsm_operations.
2892 Description: lu_conf support.
2893 Details : On a server, a file system object is uniquely identified
2894 by a fid, which is sufficient to locate and load all object
2895 state (inode). On a client, on the other hand, more data are
2896 necessary instantiate an object. Change lu_object_find() and
2897 friends to take additional `lu_conf' argument describing object.
2898 Typically this includes layout information.
2902 Description: lu_context fixes.
2903 Details : Introduce new lu_context functions that are needed on the client
2904 side, where some system threads (ptlrpcd) are shared by multiple
2905 modules, and so cannot be stopped during module shutdown.
2909 Description: Add start and stop methods to lu_device_type_operations.
2910 Details : Introduce two new methods in lu_device_type_operations, that are
2911 invoked when first instance of a given type is created and last one
2912 is destroyed respectively. This is need by CLIO.
2916 Description: Add lu_ref support to struct lu_device.
2917 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2918 track leaked references.
2922 Description: Introduce lu_kmem_descr.
2923 Details : lu_kmem_descr and its companion interface allow to create
2924 and destroy a number of kmem caches at once.
2928 Description: Fix lu_object finalization race.
2929 Details : Fix a race between lu_object_find() finding an object and its
2930 concurrent finalization. This race is (most likely) not possible
2931 on the server, but might happen on the client.
2935 Description: Introduce lu_ref interface.
2936 Details : lu_ref is a debugging module allowing to track references to
2937 a given object. It is quite cpu expensive, and has to be
2938 explicitly enabled with --enable-lu_ref. See usage description
2943 Description: Factor lu_site procfs stats into a separate function.
2944 Details : Separate lu_site stats printing code into a separate function
2945 that can be reused on a client.
2949 Description: Constify instances of struct {lu,dt,md}_device_operations.
2950 Details : Constify instances of struct {lu,dt,md}_device_operations.
2954 Description: Introduce struct md_site and move meta-data specific parts of
2955 struct lu_site here.
2956 Details : Move md-specific fields out of struct lu_site into special struct
2957 md_site, so that lu_site can be used on a client.
2961 Description: Kill mdd_lov_destroy().
2962 Details : Remove unused mdd code.
2966 Description: Add st_block checking to multistat.c.
2967 Details : Add st_block checking to multistat.c.
2971 Description: Add lu_ref support to struct obd_device.
2972 Details : Add lu_ref tracking to obd_device.
2976 Description: Kill obd_set_fail_loc().
2977 Details : Remove unused code.
2981 Description: Add special type for ptlrpc_request interpret functions.
2982 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2983 there. Actual usage will be in CLIO.
2987 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2988 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2989 doesn't work with lockdep.
2993 Description: Add rwv.c test program.
2994 Details : New testing program exercising readv(2) and writev(2) (Qian).
2998 Description: Add sendfile.c test program.
2999 Details : New testing program exercising sendfile(2) (Jay).
3003 Description: Ratelimit a message that can be very frequent.
3004 Details : Ratelimit a memory allocation failure message that can
3009 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3010 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3014 Description: Liblustre build fixes.
3015 Details : Liblustre build fixes.
3019 Description: libcfs: add cfs_{need,cond}_resched() interface.
3020 Details : libcfs: add cfs_{need,cond}_resched() definition and
3021 implementations for Linux, NT, and liblustre.
3023 Severity : enhancement
3025 Description: More exported tunables for mballoc
3026 Details : Add support for tunable preallocation window and new tunables for
3027 large/small requests
3031 Description: Detect corruption of block bitmap and checking for preallocations
3032 Details : Checks validity of on-disk block bitmap. Also it does better
3033 checking of number of applied preallocations. When corruption is
3034 found, it turns filesystem readonly to prevent further corruptions.
3038 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3039 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3040 window to 0. If it does hit cache miss, and read-pattern is still
3041 stride-io mode, does not reset the stride window, but also does not
3042 increase the stride window length in this case.
3046 Frequency : only for big-endian servers
3047 Description: Check if system is big-endian while mounting fs with extents feature
3048 Details : Mounting a filesystem with extents feature will fail on big-endian
3049 systems since ext3-based ldiskfs is not supported on big-endian
3050 systems. This can be over-riden with "bigendian_extents" mount option.
3052 Severity : enhancement
3054 Description: The root squash functionality
3055 Details : A security feature, which is to prevent users from being able
3056 to mount lustre on their desktop, run as root, and delete
3057 all of the files in the filesystem. The goal is accomplished by
3058 remapping user id (UID) and group id (GID) of the root user to
3059 a UID and GID specified by the system administartor via Lustre
3060 configuration management server (MGS). The functionality also
3061 allows to specify sets of clients for which the remapping does
3066 Description: Excessive recovery window
3067 Details : With AT enabled, the recovery window can be excessively long (6000+
3068 seconds). To address this problem, we no longer use
3069 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3070 timeout no longer depends on the service time, it is set to
3071 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3072 time via pb_service_time.
3076 Description: Watchdog triggered on MDS failover
3077 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3078 the OSTs know that the MDS "UUID" can be reused for the same export
3079 from a different NID, so we do not need to wait for the export to be
3083 Frequency : rare, only if using MMP with Linux RAID
3085 Description: MMP doesn't work with Linux RAID
3086 Details : While using HA for Lustre servers with Linux RAID, it is possible
3087 that MMP will not detect multiple mounts. To make this work we
3088 need to unplug the device queue in RAID when the MMP block is being
3089 written. Also while reading the MMP block, we should read it from
3090 disk and not the cached one.
3092 Severity : enhancement
3094 Description: open file using fid
3095 Details : A file can be opened using just its fid, like
3096 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3099 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3102 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3103 Details : add proper locking for creation of jbd_4k slab cache
3107 Description: MMP check in ext3_remount() fails without displaying any error
3108 Details : When multiple mount protection fails during remount, proper error
3111 Severity : enhancement
3113 Description: Allow stripe size to be up to 4G-64k
3114 Details : Fix math logic to allow large stripe sizes.
3118 Description: add check for >8TB ldiskfs filesystems
3119 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3120 Don't allow >8TB ldiskfs filesystems to be mounted without
3121 force_over_8tb mount option
3126 Descriptoin: OSS complains "too many blocks" and crashes
3127 Details : In filter_commitrw_write():
3128 1. lquota_chkquota();
3129 2. obdo_from_inode();
3130 3. lquota_pending_commit();
3131 "2." has changed uid and gid in this case, then caused
3132 lquota_chkquota() and lquota_pending_commit() to use
3133 different uid and gid. That is the root of the bug.
3137 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3138 Details : Detect and print out the 'stripe count = -1' case properly.
3142 Description: Out or order replies might be lost on replay
3143 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3144 request with tid smaller than one currently being added, add it
3145 to the start, not end of the list.
3147 --------------------------------------------------------------------------------
3149 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3151 * Support for kernels:
3152 2.6.5-7.283 (SLES 9),
3153 2.6.9-55.EL (RHEL 4),
3154 2.6.16.46-0.14 (SLES 10),
3155 2.6.18.8 vanilla (kernel.org)
3156 * Client support for unpatched kernels:
3157 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3158 2.6.16 - 2.6.22 vanilla (kernel.org)
3159 * Due to recently discovered recovery problems, we do not recommend
3160 using patchless RHEL 4 clients with this or any earlier release.
3161 * Recommended e2fsprogs version: 1.39.cfs8
3162 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3163 * Starting with this release, the ldiskfs backing filesystem required
3164 by Lustre is now in its own package, lustre-ldiskfs. This package
3165 should be installed. It is versioned separately from Lustre and
3166 may be released separately in future.
3170 Description: block reactivating mgc import until all deactivates complete
3171 Details : Fix race when failing back MDT/MGS to itself (testing)
3174 Frequency : at statup only
3176 Description: mds_lov_synchronize race leads to various problems
3177 Details : simultaneous MDT->OST connections at startup can cause the
3178 sync to abort, leaving the OSC in a bad state.
3180 Severity : enhancement
3182 Description: add optional extra BUILD_VERSION info
3183 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3184 to override the lustre version.
3187 Frequency : 2.6.18 servers only
3189 Description: ll_kern_mount() doesn't release the module reference
3190 Details : The ldiskfs module reference count never drops down to 0
3191 because ll_kern_mount() doesn't release the module reference.
3196 Description: server LBUG when using old ost_num_threads parameter
3197 Details : Accept the old ost_num_threads parameter but warn that it
3198 is deprecated, and fix an off-by-one error that caused an LBUG.
3203 Description: Transient SCSI error results in persistent IO issue
3204 Details : iobuf->dr_error is not reinitialized to 0 between two
3208 Frequency : sometimes when underlying device returns I/O errors
3210 Description: OSTs not going read-only during write failures
3211 Details : OSTs are not remounted read-only when the journal commit threads
3212 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3213 instead of the ext3 wrappers.
3217 Description: poor connect scaling with increasing client count
3218 Details : Don't run filter_grant_sanity_check for more than 100 exports
3219 to improve scaling for large numbers of clients.
3222 Frequency : SLES10 only
3224 Description: sanity-quota.sh quotacheck failed: rc = -22
3225 Details : Quotas cannot be enabled on SLES10.
3228 Frequency : liblustre clients only
3230 Description: getdirentries does not give error when run on compute nodes
3231 Details : getdirentries does not fail when the size specified as an argument
3232 is too small to contain at least one entry
3234 Severity : enhancement
3236 Description: Add LNET router traceability for debug purposes
3237 Details : If a checksum failure occurs with a router as part of the
3238 IO path, the NID of the last router that forwarded the bulk data
3239 is printed so it can be identified.
3244 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3245 Details : Due to a race condition, liblustre clients were occasionally
3246 evicted incorrectly.
3248 Severity : enhancement
3250 Description: lfs setstripe use optional parameters instead of postional
3253 Severity : enhancement
3255 Description: Nanosecond timestamp support for ldiskfs
3256 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3257 resolution timestamps. There is not yet support for this at
3258 the Lustre filesystem level.
3261 Frequency : during server recovery
3263 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3264 Details : request with rq_no_resend flag not awake l_wait_event if they get
3268 Frequency : nfs export on patchless client
3270 Description: connectathon hang when test nfs export over patchless client
3271 Details : Disconnected dentry cannot be found with lookup, so we do not need
3272 to unhash it or make it invalid
3275 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3276 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3277 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3282 Description: open req refcounting wrong on reconnect
3283 Details : If reconnect happened between getting open reply from server and
3284 call to mdc_set_replay_data in ll_file_open, we will schedule
3285 replay for unreferenced request that we are about to free.
3286 Subsequent close will crash in variety of ways.
3287 Check that request is still eligible for replay in
3288 mdc_set_replay_data().
3293 Description: disable writes to filesystem when reading health_check file
3294 Details : the default for reading the health_check proc file has changed
3295 to NOT do a journal transaction and write to disk, because this
3296 can cause reads of the /proc file to hang and block HA state
3297 checking on a healthy but otherwise heavily loaded system. It
3298 is possible to return to the previous behaviour during configure
3299 with --enable-health-write.
3301 Severity : enhancement
3303 Description: 64-bit inode version
3304 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3305 to the inode. This will be required for version-based recovery.
3310 Description: MDS fails to start if a duplicate client export is detected
3311 Details : in some rare cases it was possible for a client to connect to
3312 an MDS multiple times. Upon recovery the MDS would detect this
3313 and fail during startup. Handle this more gracefully.
3315 Severity : enhancement
3317 Description: Add -o localflock option to simulate old noflock
3319 Details : This will achieve local-only flock/fcntl locks
3325 Description: log_commit_thread vs filter_destroy race leads to crash
3326 Details : Take import reference before releasing llog record semaphore
3331 Description: Wrong request locking in request set processing
3332 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3333 imp_delayed_list, in this place should be used imp_lock.
3336 Frequency : when reconnection
3338 Description: Grant Leak when osc reconnect to OST
3339 Details : When osc reconnect ost, OST(filter) should check whether it
3340 should grant more space to client by comparing fed_grant and
3341 cl_avail_grant, and return the granted space to client instead
3342 of "new granted" space, because client will call osc_init_grant
3343 to update the client grant space info.
3346 Frequency : when client reconnect to OST
3348 Description: Grant Leak when osc do resend and replay bulk write
3349 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3350 bulk write request, because the grant info while be sync between
3351 OSC and OST when reconnect, and we should ignore the grant info
3352 these of resend/replay write req.
3357 Description: Grant space more than available left space sometimes.
3358 Details : When then OST is about to be full, if two bulk writing from
3359 different clients came to OST. Accord the avaliable space of the
3360 OST, the first req should be permitted, and the second one
3361 should be denied by ENOSPC. But if the seconde arrived before
3362 the first one is committed. The OST might wrongly permit second
3363 writing, which will cause grant space > available space.
3366 Frequency : when client is evicted
3368 Description: Grant might be wrongly erased when osc is evicted by OST
3369 Details : when the import is evicted by server, it will fork another
3370 thread ptlrpc_invalidate_import_thread to invalidate the
3371 import, where the grant will be set to 0. While the original
3372 thread will update the grant it got when connecting. So if
3373 the former happened latter, the grant will be wrongly errased
3374 because of this race.
3379 Description: Checking Stale with correct fid
3380 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3381 to check whether it is stale, because sometimes, we want the
3382 enqueue happened anyway, and op_data.fid2 will not be initialized.
3384 Severity : enhancement
3386 Description: update patchless client
3387 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3390 Frequency : only with 2.4 kernel
3392 Description: random memory corruption
3393 Details : size of struct ll_inode_info is to big for union inode.u and this
3394 can be cause of random memory corruption.
3399 Description: Memory leak in recovery
3400 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3401 It should also check obd_fail before fsfilt_start, otherwise if
3402 fsfilt_start return -EROFS,(failover mds during mds recovery).
3403 then the req will return with repmsg->transno = 0 and rc = EROFS.
3404 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3405 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3406 be freed no matter whether fsfilt_commit success or not.
3409 Frequency : only with huge count clients
3411 Description: Prevents from taking the superblock lock in llap_from_page for
3413 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3414 for llap_from_page call in ll_removepage prevents from taking the
3415 superblock lock for a soon died page.
3420 Description: Not check open intent error before release open handle
3421 Details : in some rare cases, the open intent error is not checked before
3422 release open handle, which may cause
3423 ASSERTION(open_req->rq_transno != 0), because it tries to release
3424 the failed open handle.
3429 Description: Set cat log bitmap only after create log success.
3430 Details : in some rare cases, the cat log bitmap is set too early. and it
3431 should be set only after create log success.
3435 Description: Accessing a block bevice can re-enable I/O when Lustre is
3436 tearing down a device.
3437 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3442 Description: service threads may hog cpus when there are a lot of requests
3443 Details : Insert cond_resched to give other threads a chance to use some CPU
3448 Description: the cat log was not initialized in recovery
3449 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3450 unlink log on mds will not be initialized until mds post
3451 recovery. And also in mds post recovery, the unlink log will
3452 initialization will be done asynchronausly, so there will be race
3453 between add unlink log and unlink log initialization.
3457 Description: brw_stats were being printed incorrectly
3458 Details : brw_stats were being printed as log2 but all of them were not
3459 recorded as log2. Also remove some code duplication arising from
3460 filter_tally_{read,write}.
3464 Frequency : rare, only in recovery.
3465 Description: ASSERTION(req->rq_type != LI_POISON) failed
3466 Details : imp_lock should be held while iterating over imp_sending_list for
3467 prevent destroy request after get timeout in ptlrpc_queue_wait.
3471 Description: replay-single.sh test 52 fails
3472 Details : A lock's skiplist need to be cleanup when it being unlinked
3473 from its resource list.
3477 Description: Short directio read returns full requested size rather than
3479 Details : Direct I/O operations should return actual amount of bytes
3480 transferred rather than requested size.
3482 Severity : enhancement
3484 Description: metadata RPC reduction (e.g. for rm performance)
3485 Details : decrease the amount of synchronous RPC between clients and servers
3486 by canceling conflicing lock before the operation on the client
3487 and packing thier handles into the main operation RPC to server.
3489 Severity : enhancement
3491 Description: Async OSC create to avoid the blocking unnecessarily.
3492 Details : If an OST has no remaining object, system will block on the
3493 creation when it needs to create a new object on this OST. Now,
3494 ways use pre-created objects when available, instead of blocking on
3495 an empty osc while others are not empty. If we must block, we
3496 block for the shortest possible period of time.
3500 Description: improve handling recoverable errors
3501 Details : if request processig with error which can be recoverable on server
3502 request should be resend, otherwise page released from cache and
3505 Severity : enhancement
3507 Description: refine locking for avoid write wrong info into lov_objid file
3508 Details : fix possible races with add new target and write/update data in
3511 --------------------------------------------------------------------------------
3513 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3518 Frequency : on some architectures
3520 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3521 Details : Uninitialized flags sometimes cause configuration commands to
3525 Frequency : patchless clients only
3527 Description: missing __iget() symbol export
3528 Details : The __iget() symbol export is missing. To avoid the need for
3529 this on patchless clients the deathrow inode reaper is turned
3530 off, and we depend on the VM to clean up old inodes. This
3531 dependency was during via the fix for bug 12181.
3533 --------------------------------------------------------------------------------
3535 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3537 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3538 INTEROPERATE with older versions automatically. In many cases a
3539 special upgrade step is needed. Please read the
3540 user documentation before upgrading any part of a live system.
3541 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3542 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3543 * WARNING: Lustre configuration and startup changes are required with
3544 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3548 Severity : enhancement
3550 Description: MountConf
3551 Details : Lustre configuration is now managed via mkfs and mount
3552 commands instead of lmc and lconf. New obd types (MGS, MGC)
3553 are added for dynamic configuration management. See
3554 https://mail.clusterfs.com/wikis/lustre/MountConf for
3557 Severity : enhancement
3559 Description: dynamic OST addition
3560 Details : OSTs can now be added to a live filesystem
3562 Severity : enhancement
3564 Description: startup order invariance
3565 Details : MDTs and OSTs can be started in any order. Clients only
3566 require the MDT to complete startup.
3568 Severity : enhancement
3570 Description: parallel, asynchronous orphan cleanup
3571 Details : orphan cleanup is now performed in separate threads for each
3572 OST, allowing parallel non-blocking operation.
3574 Severity : enhancement
3576 Description: optimized stripe assignment
3577 Details : stripe assignments are now made based on ost space available,
3578 ost previous usage, and OSS previous usage, in order to try
3579 to optimize storage space and networking resources.
3581 Severity : enhancement
3583 Description: Permanently set tunables
3584 Details : All writable /proc/fs/lustre tunables can now be permanently
3585 set on a per-server basis, at mkfs time or on a live system.
3587 Severity : enhancement
3589 Description: Lustre message v2
3590 Details : Add lustre message format v2.
3592 Severity : enhancement
3594 Description: client OST exclusion list
3595 Details : Clients can be started with a list of OSTs that should be
3596 declared "inactive" for known non-responsive OSTs.
3600 Description: SPEC SFS validation failure on NFS v2 over lustre.
3601 Details : Changes the blocksize for regular files to be 2x RPC size,
3602 and not depend on stripe size.
3604 Severity : enhancement
3606 Description: Multiple MD RPCs in flight.
3607 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3608 To avoid overly-overloading MDS, introduce a limit on number of
3609 MDS RPCs in flight for a single client and add /proc controls
3610 to adjust this limit.
3612 Severity : enhancement
3614 Description: client read/write statistics
3615 Details : Add client read/write call usage stats for performance
3616 analysis of user processes.
3617 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3618 file access. extents_stats shows chunk size distribution.
3619 extents_stats_per_process show chunk size distribution per
3622 Severity : enhancement
3624 Description: mds statistics
3625 Details : Add detailed mds operations statistics in
3626 /proc/fs/lustre/mds/*/stats.
3630 Description: Failure of copying files with lustre special EAs.
3631 Details : Client side always return success for setxattr call for lustre
3632 special xattr (currently only "trusted.lov").
3637 Description: Refcount LNET uuids
3638 Details : The global LNET uuid list grew linearly with every startup;
3639 refcount repeated list entries instead of always adding to
3642 Severity : enhancement
3644 Description: Dynamic service threads
3645 Details : Within a small range, start extra service threads
3646 automatically when the request queue builds up.
3649 Frequency : mixed-endian client/server environments
3651 Description: mixed-endian crashes
3652 Details : The new msg_v2 system had some failures in mixed-endian
3655 Severity : enhancement
3657 Description: Easy OST removal
3658 Details : OSTs can be permanently deactivated with e.g. 'lctl
3659 conf_param lustre-OST0001.osc.active=0'
3661 Severity : enhancement
3663 Description: MGS proc entries
3664 Details : Added basic proc entries for the MGS showing what filesystems
3667 Severity : enhancement
3669 Description: provide MGS failover
3670 Details : Added config lock reacquisition after MGS server failover.
3672 Severity : enhancement
3674 Description: add Linux 2.4 support
3675 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3679 Description: a large application tries to do I/O to the same resource and dies
3680 in the middle of it.
3681 Details : Check the req->rq_arrival time after the call to
3682 ost_brw_lock_get(), but before we do anything about
3683 processing it & sending the BULK transfer request. This
3684 should help move old stale pending locks off the queue as
3685 quickly as obd_timeout.
3688 Frequency : when an incorrect nid is specified during startup
3690 Description: ptlrpc connect to non-existant node causes kernel crash
3691 Details : LNET can't be re-entered from an event callback, which
3692 happened when we expire a message after the export has been
3693 cleaned up. Instead, hand the zombie cleanup off to another
3696 Severity : enhancement
3698 Description: plain/inodebits lock performance improvement
3699 Details : Grouping plain/inodebits in granted list by their request modes
3700 and bits policy, thus improving the performance of search through
3704 Frequency : only if OST filesystem is corrupted
3706 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3707 Details : for a short time RPCs with bulk IO are in the replay list,
3708 but replay of bulk IOs is unimplemented. If the OST filesystem
3709 is corrupted due to disk cache incoherency and then replay is
3710 started it is possible to trip an assertion. Avoid putting
3711 committed RPCs into the replay list at all to avoid this issue.
3714 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3716 Description: System hang on startup
3717 Details : This bug allowed the liblustre (e.g. catamount) client to
3718 return to the app before handling all startup RPCs. This
3719 could leave the node unresponsive to lustre network traffic
3720 and manifested as a server ptllnd timeout.
3722 Severity : enhancement
3724 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3725 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3726 boolean to control whether to print peer debug info when a
3727 client's RPC times out.
3730 Frequency : only for kernels with patches from Lustre below 1.4.3
3732 Description: Remove old rdonly API
3733 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3736 Frequency : only for devices with external journals
3738 Description: Set external device read-only also
3739 Details : During a commanded failover stop, we set the disk device
3740 read-only while the server shuts down. We now also set any
3741 external journal device read-only at the same time.
3744 Frequency : when upgrading from 1.4 while trying to change parameters
3746 Description: The wrong (new) MDC name was used when setting parameters for
3747 upgraded MDT's. Also allows changing of OSC (and MDC)
3748 parameters if --writeconf is specified at tunefs upgrade time.
3751 Frequency : when setting specific ost indicies
3753 Description: QOS code breaks on skipped indicies
3754 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3755 created with --index need not be sequential.
3757 Severity : enhancement
3759 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3760 Details : The uninit_groups feature works in conjunction with the kernel
3761 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3762 up the pass1 processing of e2fsck. This is a read-only feature
3763 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3764 mount filesystems that have had this feature enabled.
3766 Severity : enhancement
3768 Description: Improve multi-block allocation algorithm to avoid fragmentation
3769 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3770 allocation locality and avoid filesystem fragmentation.
3772 ------------------------------------------------------------------------------
3774 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3776 * Support for kernels:
3777 2.4.21-47.0.1.EL (RHEL 3)
3778 2.6.5-7.283 (SLES 9)
3779 2.6.9-42.0.10.EL (RHEL 4)
3780 2.6.12.6 vanilla (kernel.org)
3781 2.6.16.27-0.9 (SLES 10)
3782 * Recommended e2fsprogs version: 1.39.cfs5
3784 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3788 Frequency : occasional, depends on client load and configuration
3789 Bugzilla : 12181, 12203
3790 Description: data loss for recently-modified files
3792 Details : In some cases it is possible that recently written or created
3793 files may not be written to disk in a timely manner (this should
3794 normally be within 30s unless client IO load is very high).
3795 The problem appears as zero-length files or files that are a
3796 multiple of 1MB in size after a client crash or client eviction
3797 that are missing data at the end of the file.
3799 This problem is more likely to be hit on clients where files are
3800 repeatedly created and unlinked in the same directory, clients
3801 have a large amount of RAM, have many CPUs, the filesystem has
3802 many OSTs, the clients are rebooted frequently, and/or the files
3803 are not accessed by other nodes after being written.
3805 The presence of the problem can be detected by looking at
3806 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3807 smaller than the second (nr_unused) then dirty files will not
3808 be flushed automatically to disk. "sync; sleep 10" should be
3809 run several times on the node before unmounting it to update
3810 Lustre (this is also safe to run on nodes without this problem).
3812 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3813 that can cause this same problem, so customers using that kernel
3814 also need to update the kernel in addition to Lustre. In order
3815 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3817 It is normal that files written just before a client crash (less
3818 than 30s) may not yet have been flushed to disk, even for local
3822 Frequency : frequent on thin XT3 nodes
3824 Description: UUID collision on thin XT3 Linux nodes
3825 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3826 randomly, since we relied on an insufficiently-seeded PRNG.
3831 Description: OSS hangs after "All ost request buffers busy"
3832 Details : A deadlock between quota and journal operations caused OSS
3833 hangs after printing "All ost request buffers busy."
3836 Frequency : always on liblustre builds
3838 Description: Cleanup compiler warnings on liblustre
3841 Frequency : always on liblustre builds on XT3
3843 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3847 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3848 > 32000 subdirectories
3849 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3850 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3851 subdirectory count crosses 32000. This will aid e2fsck to
3852 correctly handle more than 32000 subdirectories.
3855 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3857 Description: System hang on startup
3858 Details : This bug allowed the liblustre (e.g. catamount) client to
3859 return to the app before handling all startup RPCs. This
3860 could leave the node unresponsive to lustre network traffic
3861 and manifested as a server ptllnd timeout.
3863 Severity : enhancement
3865 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3866 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3867 boolean to control whether to print peer debug info when a
3868 client's RPC times out.
3873 Description: make O_SYNC working on 2.6 kernels
3874 Details : 2.6 kernels use different method for mark pages for write,
3875 so need add a code to lustre for O_SYNC work.
3880 Description: Failure to close file and release space on NFS
3881 Details : Put inode details into lock acquired in ll_intent_file_open.
3882 Use mdc_intent_lock in ll_intent_open to properly
3883 detect all kind of errors unhandled by mdc_enqueue.
3888 Description: proc file read during shutdown sometimes raced obd removal,
3890 Details : Add lock to prevent obd access after proc file removal.
3893 Frequency : Only for files larger than 4GB on 32-bit clients.
3895 Description: improperly doing page alignment of locks
3896 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3897 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3902 Description: RPCs being resent when they shouldn't be
3903 Details : Some RPCs that should not be resent are being resent. This
3904 can cause inconsistencies in the RPC state machine. Do not
3905 resend such requests.
3908 Frequency : rare, only with NFS export
3910 Description: Crash on NFS re-export node
3911 Details : under very unusual load conditions an assertion is hit in
3912 ll_intent_file_open()
3915 Frequency : only if OST filesystem is corrupted
3917 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3918 Details : for a short time RPCs with bulk IO are in the replay list,
3919 but replay of bulk IOs is unimplemented. If the OST filesystem
3920 is corrupted due to disk cache incoherency and then replay is
3921 started it is possible to trip an assertion. Avoid putting
3922 committed RPCs into the replay list at all to avoid this issue.
3927 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3928 Details : Large single O_DIRECT read and write calls can fail to allocate
3929 a sufficiently large buffer to process the request. In case of
3930 allocation failure the allocation is retried with a smaller
3931 buffer and broken into smaller requests.
3933 Severity : enhancement
3935 Description: Add -o localflock option to simulate old noflock behaviour.
3936 Details : This will achieve local-only flock/fcntl locks coherentness.
3941 Description: versioning check is incomplete
3942 Details : Checking the version difference of client vs. server, report
3943 error if the gap is too big.
3948 Description: add support PG_writeback bit
3949 Details : add support for PG_writeback bit for Lustre, for more carefull
3950 work with page cache in 2.6 kernel. This also fix some deadlocks
3951 and remove hack for work O_SYNC with 2.6 kernel.
3953 Severity : enhancement
3955 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3956 Details : The uninit_groups feature works in conjunction with the kernel
3957 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3958 up the pass1 processing of e2fsck. This is a read-only feature
3959 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3960 mount filesystems that have had this feature enabled.
3962 Severity : enhancement
3964 Description: Improve multi-block allocation algorithm to avoid fragmentation
3965 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3966 allocation locality and avoid filesystem fragmentation.
3968 ------------------------------------------------------------------------------
3970 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3972 * Support for kernels:
3973 2.6.9-42.0.3.EL (RHEL 4)
3974 2.6.5-7.276 (SLES 9)
3975 2.4.21-47.0.1.EL (RHEL 3)
3976 2.6.12.6 vanilla (kernel.org)
3977 2.6.16.21-0.8 (SLES10)
3978 * Recommended e2fsprogs version: 1.39.cfs2-0
3980 * The backwards-compatible /proc/sys/portals symlink has been removed
3981 in this release. Before upgrading, please ensure that you change
3982 any configuration scripts or /etc/sysctl.conf files that access
3983 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3984 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3985 in advance of the upgrade on any system running Lustre 1.4.6 or
3986 newer, since /proc/sys/lnet was added in that version.
3987 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3991 Frequency : only when quota is used
3993 Description: avoid scanning export list for quota master
3994 Details : Change the algorithms to avoid scanning export list in order
3995 to improve the efficiency.
3998 Frequency : MDS failover only, very rarely
4000 Description: "went back in time" messages on mds failover
4001 Details : The greatest transno may be lost when the current operation
4002 finishes with an error (transno==0) and the client's last_rcvd
4003 record is over-written. Save the greatest transno in the
4004 mds_last_transno for this case.
4007 Frequency : always for specific kernels and striping counts
4009 Description: client may get "Matching packet too big" without ACL support
4010 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4011 when trying to access files in certain configurations. The
4012 clients should in fact be denied when mounting because they do
4013 not understand ACLs.
4016 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4018 Description: many clients connecting with IO in progress causes connect timeouts
4019 Details : Avoid synchronous journal commits to avoid delays caused by many
4020 clients connecting/disconnecting when bulk IO is in progress.
4021 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4022 OST_IO_PORTAL to avoid delays behind potentially many pending
4026 Frequency : occasionally with multiple writers to a single file
4028 Description: shared writes to file may result in wrong size reported by stat()
4029 Details : Allow growing of kms when extent lock is cancelled
4032 Frequency : always with random mmap IO to multi-striped file
4034 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4035 Details : Only if the hole is at the end of OST object so that kms is too
4036 small. Fix is to increase kms accordingly in ll_nopage.
4039 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4041 Description: writes to a missing object would leak memory on the OST
4042 Details : If there is an inconsistency between the MDS and OST filesystems,
4043 such that the MDS references an object that doesn't exist, writes
4044 to that object will leak memory due to incorrect cleanup in the
4045 error handling path, eventually running out of memory on the OST.
4050 Description: Creating too long symlink causes lustre errors
4051 Details : Check symlink and name lengths before sending requests to MDS.
4054 Frequency : only if flock is enabled (not on by default)
4056 Description: posix locks not released on fd closure on 2.6.9+
4057 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4058 kernels, this caused such locks not to be released on fd close and
4059 then assertions on fs unmount about still used locks.
4062 Frequency : MDS failover only, very rarely
4064 Description: clients may get ASSERTION(granted_lock != NULL)
4065 Details : When request was taking a long time, and a client was resending
4066 a getattr by name lock request. The were multiple lock requests
4067 with the same client lock handle and
4068 mds_getattr_name->fixup_handle_for_resent_request found one of the
4069 lock handles but later failed with ASSERTION(granted_lock != NULL).
4074 Description: handle->h_buffer_credits > 0, assertion failure
4075 Details : h_buffer_credits is zero after truncate, causing assertion
4076 failure. This patch extends the transaction or creates a new
4080 Frequency : NFS re-export or patchless client
4081 Bugzilla : 11179, 10796
4082 Description: Crash on NFS re-export node (__d_move)
4083 Details : We do not want to hash the dentry if we don't have a lock.
4084 But if this dentry is later used in d_move, we'd hit uninitialised
4085 list head d_hash, so we just do this to init d_hash field but
4086 leave dentry unhashed.
4089 Frequency : NFS re-export or patchless client
4091 Description: NFS exports has problem with symbolic link
4092 Details : lustre client didn't properly install dentry when re-exported
4093 to NFS or running patchless client.
4096 Frequency : NFS re-export or patchless client
4098 Description: Various nfs/patchless fixes.
4099 Details : fixes reuse disconected alias for lookup process - this fixes
4100 warning "find_exported_dentry: npd != pd",
4101 fix permission error with open files at nfs.
4102 fix apply umask when do revalidate.
4105 Frequency : occasional
4107 Description: Crash on NFS re-export node
4108 Details : calling clear_page() on the wrong pointer triggered oops in
4109 generic_mapping_read().
4112 Frequency : rarely, using O_DIRECT IO
4114 Description: unaligned directio crashes client with LASSERT
4115 Details : check for unaligned buffers before trying any requests.
4118 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4120 Description: stale data returned from RAID cache
4121 Details : If only a small amount of IO is done to the RAID device before
4122 reading it again it is possible to get stale data from the RAID
4123 cache instead of reading it from disk.
4126 Frequency : always for sles10 kernel
4128 Description: sles10 support
4129 Details : ll_follow_link: compile fixes and using of nd_set_link
4130 under newer kernels.
4133 Frequency : depends on arch, kernel and compiler version, always on sles10
4136 Description: recursive or deep enough symlinks cause stack overflow
4137 Details : getting rid of large stack-allocated variable in
4141 Frequency : depends on hardware
4143 Description: lustre write performance loss in the SLES10 kernel
4144 Details : the performance loss is caused by using of write barriers in the
4145 ext3 code. The SLES10 kernel turns barrier support on by
4146 default. The fix is to undo that change for ldiskfs.
4148 ------------------------------------------------------------------------------
4150 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4152 * Support for kernels:
4153 2.6.9-42.0.3EL (RHEL 4)
4154 2.6.5-7.276 (SLES 9)
4155 2.4.21-47.0.1.EL (RHEL 3)
4156 2.6.12.6 vanilla (kernel.org)
4160 Frequency : quota enabled and large files being deleted
4162 Description: releasing more than 4GB of quota at once hangs OST
4163 Details : If a user deletes more than 4GB of files on a single OST it
4164 will cause the OST to spin in an infinite loop. Release
4165 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4170 Description: statfs data retrieved from /proc may be stale or zero
4171 Details : When reading per-device statfs data from /proc, in the
4172 {kbytes,files}_{total,free,avail} files, it may appear
4173 as zero or be out of date.
4176 Frequency : systems with MD RAID1 external journal devices
4178 Description: lconf's call to blkid is confused by RAID1 journal devices
4179 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4180 of returning all block devices that match the journal UUID.
4183 Frequency : always, for aggregate stripe size over 4GB
4185 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4186 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4187 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4190 Frequency : always if "lfs find" used on a local file/directory
4192 Description: "lfs find" segfaults if used on a local file/directory
4193 Details : The case where a directory component was not specified wasn't
4194 handled correctly. Handle this properly.
4197 Frequency : always on ppc64
4199 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4200 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4201 perform endianness conversion.
4204 Frequency : rarely (truncate to non-zero file size after write under load)
4205 Bugzilla : 10730, 10687
4206 Description: Files padded with zeros to next 4K multiple
4207 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4208 it is possible that files that are truncated to a non-zero size
4209 immediately after being written are filled with zero bytes beyond
4210 the truncated size. No file data is lost.
4212 Severity : enhancement
4214 Description: Allow recovery/failover for liblustre clients.
4215 Details : liblustre clients were unaware of failover configurations until
4218 Severity : enhancement
4220 Description: user file locks should fail when not mounting with flock option
4221 Details : Set up an error-returning stub in ll_file_operations.lock field
4222 to prevent incorrect behaviour when client is mounted without
4223 flock option. Also, set up properly f_op->flock field for
4227 Frequency : always on ia64
4229 Description: "lfs df" loops on printing out MDS statfs information
4230 Details : The obd_ioctl_data was not initialized and in some systems
4231 this caused a failure during the ioctl that did not return
4232 an error. Initialize the struct and return an error on failure.
4235 Frequency : SLES 9 only
4237 Description: Error of copying files with lustre special EAs as root
4238 Details : Client side always return success for setxattr call for lustre
4239 special xattr (currently only "trusted.lov").
4242 Frequency : rarely on clusters with both ia64+i386 clients
4244 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4245 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4246 clients are doing concurrent writes to the same file it is
4247 possible that the ia64 clients may LASSERT because the OST
4248 extent locks are not PAGE_SIZE aligned. Ensure that grown
4249 locks are always aligned on the request boundary.
4252 Frequency : specific use, occasional
4254 Description: Overwriting in use executable truncates on-disk binary image
4255 Details : If one node attempts to overwrite an executable in use by
4256 another node, we now correctly return ETXTBSY instead of
4257 truncating the file.
4259 Severity : enhancement
4261 Description: Async OSC create to avoid the blocking unnecessarily.
4262 Details : If an OST has no remaining object, system will block on the creating
4263 when need to create a new object on this OST. Now, ways use
4264 pre-created objects when available, instead of blocking on an
4265 empty osc while others are not empty. If we must block, we block
4266 for the shortest possible period of time.
4271 Description: chmod on Lustre root is propagated to other clients
4272 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4273 invalid by the follow_mount time.
4276 Frequency : liblustre clients only
4278 Description: Race in 'instant cancel' lock handling could lead to such locks
4279 never to be granted in case of SMP MDS
4280 Details : Do not destroy not yet granted but cbpending locks in
4284 Frequency : replay/resend of open
4286 Description: non null lock assetion failure in mds_intent_policy
4287 Details : Trying to replay/resend lockless open requests resulted in
4288 mds_open() returning 0 with no lock. Now it sets a flag if
4289 a lock is going to be returned.
4291 Severity : enhancement
4293 Description: Checksum enhancements
4294 Details : New checksum enhancements allow for resending RPCs that failed
4297 Severity : enhancement
4299 Description: Tunables on number of dirty pages in cacche
4300 Details : Allow to set limit on number of dirty pages cached.
4305 Description: client crash on unmount - lock still has references
4306 Details : In some error handling cases it was possible to leak a lock
4307 reference on a client while accessing a file. This was not
4308 harmful to the client during operation, but would cause the
4309 client to crash when the filesystem is unmounted.
4312 Frequency : specific case, rare
4314 Description: ETXTBSY on mds though file not in use by client
4315 Details : ETXTBSY is no longer incorrectly returned when attempting to
4316 chmod or chown a directory that the user previously tried to
4317 execute or a currently-executing binary.
4320 Frequency : extremely rare except on liblustre-based clients
4322 Description: Lustre space not freed when files are deleted
4323 Details : Clean up open-unlinked files after client eviction. Previously
4324 the unlink was skipped and the files remained as orphans.
4329 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4330 Details : In some cases it was possible to send a blocking callback to
4331 a client doing a glimpse, even though that client didn't get
4332 a lock granted. When the glimpse lock is cancelled on the OST
4333 the freed lock is left on the waiting list and corrupted the list.
4336 Frequency : all core dumps
4338 Description: Broke core dumps to lustre
4339 Details : Negative dentry may be unhashed if parent does not have UPDATE
4340 lock, but some callers, e.g. do_coredump, expect dentry to be
4341 hashed after successful create, hash it in ll_create_it.
4343 ------------------------------------------------------------------------------
4345 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4347 * Support for kernels:
4348 2.6.9-42.0.2.EL (RHEL 4)
4349 2.6.5-7.276 (SLES 9)
4350 2.4.21-40.EL (RHEL 3)
4351 2.6.12.6 vanilla (kernel.org)
4355 Frequency : always on RHEL 3
4357 Description: Number of open files grows over time
4358 Details : The number of open files grows over time, whether or not
4359 Lustre is started. This was due to a filp leak introduced
4360 by one of our kernel patches.
4362 ------------------------------------------------------------------------------
4364 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4366 * Support for kernels:
4367 2.6.9-42.EL (RHEL 4)
4368 2.6.5-7.276 (SLES 9)
4369 2.4.21-40.EL (RHEL 3)
4370 2.6.12.6 vanilla (kernel.org)
4375 Bugzilla : 5719, 9635, 9792, 9684
4376 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4377 Details : If a server is under heavy load and cannot reply to new
4378 connection requests before the client resends the (re)connect,
4379 the connection handling code can behave badly if two service
4380 threads are concurrently handing separate (re)connections from
4381 the same client. Add better locking to the connection handling
4382 code, and ensure that only a single connection will be processed
4383 for a given client UUID, even if the lock is dropped.
4385 Severity : enhancement
4387 Description: add TCP zero-copy support to kernel
4388 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4389 sends if the hardware supports scatter-gather and checksumming.
4390 This allows socklnd to do client-write and server-read more
4391 efficiently and reduce CPU utilization from skbuf copying.
4394 Frequency : only if NFS exporting from client
4396 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4397 Details : If setting attributes on a file created under NFS that had
4398 never been opened it would be possible to oops the client
4399 if the file had no objects.
4402 Frequency : always for liblustre
4404 Description: liblustre client does MDS+OSTs setattr RPC for each write
4405 Details : When doing a write from a liblustre client, the client
4406 incorrectly issued an RPC to the MDS and each OST the file was
4407 striped over in order to update the timestamps. When writing
4408 with small chunks and many clients this could overwhelm the MDS
4409 with RPCs. In all cases it would slow down the write because
4410 these RPCs are unnecessary.
4412 Severity : enhancement
4414 Description: allow number of MDS service threads to be changed at module load
4415 Details : It is now possible to change the number of MDS service threads
4416 running. Adding "options mds mds_num_threads={N}" to the MDS's
4417 /etc/modprobe.conf will set the number of threads for the next
4418 time Lustre is restarted (assuming the "mds" module is also
4419 reloaded at that time). The default number of threads will
4420 stay the same, 32 for most systems.
4425 Description: OST crash if filesystem is unformatted or corrupt
4426 Details : If an OST is started on a device that has never been formatted
4427 or if the filesystem is corrupt and cannot even mount then the
4428 error handling cleanup routines would dereference a NULL pointer.
4433 Description: NULL pointer deref in llap_from_page.
4434 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4435 incorrect) mapping if the page was truncated/reclaimed while it was
4436 searched for. Check for this condition and skip such pages when
4437 doing readahead. Introduce extra check to llap_from_page() to
4438 verify page->mapping->host is non-NULL (so page is not anonymous).
4441 Frequency : Sometimes when using sys_sendfile
4443 Description: "page not covered by a lock" warnings from ll_readpage
4444 Details : sendfile called ll_readpage without right page locks present.
4445 Now we introduced ll_file_sendfile that does necessary locking
4446 around call to generic_file_sendfile() much like we do in
4450 Frequency : with certain MDS communication failures at client mount time
4452 Description: NULL pointer deref after failed client mount
4453 Details : a client connection request may delayed by the network layer
4454 and not be sent until after the PTLRPC layer has timed out the
4455 request. If the client fails the mount immediately it will try
4456 to clean up before the network times out the request. Add a
4457 reference from the request import to the obd device and delay
4458 the cleanup until the network drops the request.
4461 Frequency : occasionally during client (re)connect
4463 Description: assertion failure during client (re)connect
4464 Details : processing a client connection request may be delayed by the
4465 client or server longer than the client connect timeout. This
4466 causes the client to resend the connection request. If the
4467 original connection request is replied in this interval, the
4468 client may trip an assertion failure in ptlrpc_connect_interpret()
4469 which thought it would be the only running connect process.
4472 Frequency : only with obd_echo servers and clients that are rebooted
4474 Description: kernel BUG accessing uninitialized data structure
4475 Details : When running an obd_echo server it did not start the ping_evictor
4476 thread, and when a client was evicted an uninitialized data
4477 structure was accessed. Start the ping_evictor in the RPC
4478 service startup instead of the OBD startup.
4480 Severity : enhancement
4481 Bugzilla : 10193 (patchless)
4482 Description: Remove dependency on various unexported kernel interfaces.
4483 Details : No longer need reparent_to_init, exit_mm, exit_files,
4484 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4487 Frequency : rare (only users of deprecated and unsupported LDAP config)
4489 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4490 Details : LDAP apparently contains 'lustreName' attributes instead of
4491 'name'. A simple remapping of the name is sufficient.
4494 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4496 Description: waiting_locks_callback trips kernel BUG if client is evicted
4497 Details : Running with the dump_on_timeout debug flag turned on makes
4498 it possible that the waiting_locks_callback() can try to dump
4499 the Lustre kernel debug logs from an interrupt handler. Defer
4500 this log dumping to the expired_lock_main() thread.
4502 Severity : enhancement
4504 Description: Support NFS exporting on 2.6 kernels.
4505 Details : Implement non-rawops metadata methods for NFS server to use without
4506 changing NFS server code.
4509 Frequency : very rare (synthetic metadata workload only)
4511 Description: two racing renames might cause an MDS thread to deadlock
4512 Details : Running the "racer" program may cause one MDS thread to rename
4513 a file from being the source of a rename to being the target of
4514 a rename at exactly the same time that another thread is doing
4515 so, and the second thread has already enqueued these locks after
4516 doing a lookup of the target and is trying to relock them in
4517 order. Ensure that we don't try to re-lock the same resource.
4520 Frequency : only very large systems with liblustre clients
4522 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4523 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4524 all OSTs in parallel. This allows the eviction of stale liblustre
4525 clients to proceed much faster than if they were done in series,
4526 and also offers similar improvements for other set_info RPCs.
4531 Description: excessive CPU usage during initial read phase on client
4532 Details : During the initial read phase on a client, it would agressively
4533 retry readahead on the file, consuming too much CPU and impacting
4534 performance (since 1.4.5.8). Improve the readahead algorithm
4535 to avoid this, and also improve some other common cases (read
4536 of small files in particular, where "small" is files smaller than
4537 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4542 Description: MDS crash when receiving packet with unknown intent.
4543 Details : Do not LBUG in unknown intent case, just return -EFAULT
4545 Severity : enhancement
4546 Bugzilla : 9293, 9385
4547 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4548 Details : Do not serialize getattr (non-intent version) and statfs.
4551 Frequency : occasional, when OST network is overloaded/intermittent
4553 Description: client evicted by OST after bulk IO timeout
4554 Details : If a client sends a bulk IO request (read or write) the OST
4555 may evict the client if it is unresposive to its data GET/PUT
4556 request. This is incorrect if the network is overloaded (takes
4557 too long to transfer the RPC data) or dropped the OST GET/PUT
4558 request. There is no need to evict the client at all, since
4559 the pinger and/or lock callbacks will handle this, and the
4560 client can restart the bulk request.
4563 Frequency : Always when mmapping file with no objects
4565 Description: client crashes when mmapping file with no objects
4566 Details : Check that we actually have objects in a file before doing any
4567 operations on objects in ll_vm_open, ll_vm_close and
4573 Description: Request leak when working with deleted CWD
4574 Details : Introduce advanced request refcount tracking for requests
4575 referenced from lustre intent.
4577 Severity : Enhancement
4579 Description: Cache open file handles on client.
4580 Details : MDS now will return special lock along with openhandle, if
4581 requested and client is allowed to hold openhandle, even if unused,
4582 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4583 closing files for every read/write openration.
4585 Severity : Enhancement
4587 Description: Cache open negative dentries on client when possible.
4588 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4589 negative dentries on lock revocation.
4594 Description: Remounting a client read-only wasn't possible with a zconf mount
4595 Details : It wasn't possible to remount a client read-only with llmount.
4597 Severity : enhancement
4598 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4599 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4600 MPICH is included to add Lustre-specific ADIO interfaces.
4601 This is based closely on the UFS ADIO layer and only differs
4602 in file creation, in order to allow the OST striping to be set.
4603 This is user-contributed code and not supported by CFS.
4608 Description: extended inode attributes (immutable, append-only) work improperly
4609 when 2.4 and 2.6 kernels are used on client/server or vice versa
4610 Details : Introduce kernel-independent values for these flags.
4612 Severity : enhancement
4615 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4616 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4617 granularity. Now, Fractional MB granularity is be supported,
4618 this is very useful for low memory system.
4620 Severity : enhancement
4622 Description: Getattr by fid
4623 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4624 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4625 have locks and avoids getattr rpc requests for it on every stat.
4628 Frequency : Always, for filesystems larger than 2TB
4630 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4631 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4632 filesystems larger than 2TB to be mounted with mballoc enabled.
4635 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4637 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4638 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4639 filesystems larger than 2TB is mounted then the kernel will
4640 silently corrupt the start of the filesystem. CONFIG_LBD is
4641 enabled for all CFS-supported kernels, but the possibility of
4642 this happening with a modified kernel config exists.
4644 Severity : enhancement
4646 Description: add client O_DIRECT support for 2.6 kernels
4647 Details : It is now possible to do O_DIRECT reads and writes to files
4648 in the Lustre client mountpoint on 2.6 kernel clients.
4650 Severity : enhancement
4652 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4653 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4654 parallel, not waiting for response from every OST before sending
4655 a rpc to the next OST.
4660 Description: setattr vs write race when updating file timestamps
4661 Details : Client processes that update a file timestamp into the past
4662 right after writing to the file (e.g. tar) it is possible that
4663 the updated file modification time can be reset to the current
4664 time due to a race between processing the setattr and write RPC.
4666 Severity : enhancement
4668 Description: Bring 'lfs find' closer in line with regular Linux find.
4669 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4670 -print0 options and obtains all the needed info through the lustre
4673 Severity : enhancement
4675 Description: support up to 1024 configured devices on one node
4676 Details : change obd_dev array from statically allocated to dynamically
4677 allocated structs as they are first used to reduce memory usage
4682 Description: Flush dirty partially truncated pages during truncate
4683 Details : Immediatelly flush partially truncated pages in filter_setattr,
4684 this way we completely avoid having any pages in page cache on OST
4685 and can retire ugly workarounds during writes to flush such pages.
4690 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4691 Details : protect inode from truncation within vfs_unlink() context
4692 just take a reference before calling vfs_unlink() and release it
4693 when parent's i_sem is free.
4698 Description: last_id value checked outside lock on OST caused LASSERT failure
4699 Details : If there were multiple MDS->OST object precreate requests in
4700 flight, it was possible that the OST's last object id was checked
4701 outside a lock and incorrectly tripped an assertion. Move checks
4702 inside locks, and discard old precreate requests.
4705 Frequency : always, if extents are used on OSTs
4707 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4708 Details : OSTs using the extents format would not zero the high 16 bits of
4709 the index physical block number. This is not a problem for any
4710 OST filesystems smaller than 16TB, and no kernels support ext3
4711 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4712 new/modified files) and can be fixed for existing filesystems
4713 with e2fsprogs-1.39-cfs1.
4718 Description: import connection selection may be incorrect if timer wraps
4719 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4720 import connections to be ignored if the 32-bit jiffies counter
4721 wraps. Use a 64-bit jiffies counter.
4724 Frequency : very large clusters immediately after boot
4726 Description: LNET request buffers exhausted under heavy short-term load
4727 Details : If a large number of client requests are generated on a service
4728 that has previously never seen so many requests it is possible
4729 that the request buffer growth cannot keep up with the spike in
4730 demand. Instead of dropping incoming requests, they are held in
4731 the LND until the RPC service can accept more requests.
4734 Frequency : Sometimes during replay
4736 Description: Assertion failure in ll_local_open after replay.
4737 Details : If replay happened on an open request reply before we were able
4738 to set replay handler, reply will become not swabbed tripping the
4739 assertion in ll_local_open. Now we set the handler right after
4740 recognising of open request
4743 Frequency : very rare
4745 Description: kernel reports "badness in vsnprintf"
4746 Details : Reading from the "recovery_status" /proc file in small chunks
4747 may cause a negative length in lprocfs_obd_rd_recovery_status()
4748 call to vsnprintf() (which is otherwise harmless). Exit early
4749 if there is no more space in the output buffer.
4751 Severity : enhancement
4753 Description: clear OBD RPC statistics by writing to them
4754 Details : It is now possible to clear the OBD RPC statistics by writing
4755 to the "stats" file.
4760 Description: Client mtime is not the same on different clients after utimes
4761 Details : In some cases, the client was using the utimes() syscall on
4762 a file cached on another node. The clients now validate the
4763 ctime from the MDS + OSTs to determine which one is right.
4768 Description: Inability to activate failout mode
4769 Details : lconf script incorrectly assumed that in pythong string's numeric
4770 value is used in comparisons.
4773 Frequency : always with multiple stripes per file
4775 Description: Inefficient object allocation for mutli-stripe files
4776 Details : When selecting which OSTs to stripe files over, for files with
4777 a stripe count that divides evenly into the number of OSTs,
4778 the MDS is always picking the same starting OST for each file.
4779 Return the OST selection heuristic to the original design.
4784 Description: mount failures may take full timeout to return an error
4785 Details : Under some heavy load conditions it is possible that a
4786 failed mount can wait for the full obd_timeout interval,
4787 possibly several minutes, before reporting an error.
4788 Instead return an error as soon as the status is known.
4790 Frequency : quota enabled and large files being deleted
4792 Description: releasing more than 4GB of quota at once hangs OST
4793 Details : If a user deletes more than 4GB of files on a single OST it
4794 will cause the OST to spin in an infinite loop. Release
4795 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4800 Description: statfs data retrieved from /proc may be stale or zero
4801 Details : When reading per-device statfs data from /proc, in the
4802 {kbytes,files}_{total,free,avail} files, it may appear
4803 as zero or be out of date.
4806 Frequency : systems with MD RAID1 external journal devices
4808 Description: lconf's call to blkid is confused by RAID1 journal devices
4809 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4810 of returning all block devices that match the journal UUID.
4813 Frequency : always, for aggregate stripe size over 4GB
4815 Description: assertion fails when trying to use 4GB stripe size
4816 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4817 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4820 Frequency : always on ppc64
4822 Description: the first write on an ext3 filesystem with mballoc got stuck
4823 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4824 perform endianness conversion.
4826 ------------------------------------------------------------------------------
4828 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4830 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4831 INTEROPERATE with older versions automatically. Please read the
4832 user documentation before upgrading any part of a live system.
4833 * WARNING: Lustre networking configuration changes are required with
4834 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4837 * Support for kernels:
4838 2.6.9-22.0.2.EL (RHEL 4)
4839 2.6.5-7.244 (SLES 9)
4840 2.6.12.6 vanilla (kernel.org)
4843 Severity : enhancement
4844 Bugzilla : 7981/8208
4845 Description: Introduced Lustre Networking (LNET)
4846 Details : LNET is new networking infrastructure for Lustre, it includes
4847 a reorganized network configuration mode (see the user
4848 documentation for full details) as well as support for routing
4849 between different network fabrics. Lustre Networking Devices
4850 (LNDs) for the supported network fabrics have also been
4851 created for this new infrastructure.
4853 Severity : enhancement
4854 Description: Introduced Access control lists
4855 Details : clients can set ACLs on files and directories in order to have
4856 more fine-grained permissions than the standard Unix UGO+RWX.
4857 The MDS must be started with the "-o acl" mount option.
4859 Severity : enhancement
4860 Description: Introduced filesystem quotas
4861 Details : Administrators may now establish per-user quotas on the
4864 Severity : enhancement
4866 Description: Configuration change for the XT3
4867 The PTLLND is now used to run Lustre over Portals on the XT3
4868 The configure option(s) --with-cray-portals are no longer used.
4869 Rather --with-portals=<path-to-portals-includes> is used to
4870 enable building on the XT3. In addition to enable XT3 specific
4871 features the option --enable-cray-xt3 must be used.
4876 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4877 Details : A race between allocating a new llcd and re-getting the llcd_lock
4878 allowed another thread to grab newly-allocated llcd.
4880 Severity : enhancement
4882 Description: 2.6 OST async journal commit and locking fix to improve performance
4883 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4884 2.6 kernels are now async as they already were in 2.4 kernels so
4885 that they can commit concurrently with the network bulk transfer.
4886 For block-allocated files the filter allocation semaphore is held
4887 to avoid filesystem fragmentation during allocation. BKL lock
4888 removed for 2.6 xattr operations where it is no longer needed.
4893 Description: lconf incorrectly determined whether two IP networks could talk
4894 Details : In some more complicated routing and multiple-network
4895 configurations, lconf will avoid trying to make a network
4896 connection to a disjoint part of the IP space. It was doing the
4897 math incorrectly for one set of cases.
4902 Description: Fix for potential infinite loop processing records in an llog.
4903 Details : If an llog record is corrupted/zeroed, it is possible to loop
4904 forever in llog_process(). Validate the llog record length
4905 and skip the remainder of the block on error.
4908 Frequency : occasional (liblustre only)
4910 Description: liblustre could not open files whose last component is a symlink
4911 Details : sysio_path_walk() would incorrectly pass the open intent to
4912 intermediate path components.
4915 Frequency : rare (liblustre only with non-standard tuning)
4916 Bugzilla : 7201 (7350)
4917 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4918 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4919 after it has been released, only do this for referenced locks
4921 Severity : enhancement
4923 Description: specifying an (invalid) directory default stripe_size of -1
4924 would reset the directory default striping
4925 Details : stripe_size -1 was used internally to signal directory stripe
4926 removal, now use "all default" to signal dir stripe removal
4927 as a directory striping of "all default" is not useful
4930 Frequency : common for large clusters running liblustre clients
4932 Description: doing an ls when liblustre clients are running is slow
4933 Details : sending a glimpse AST to a liblustre client waits for every AST
4934 to time out, as liblustre clients will not respond. Since they
4935 cannot cache data we refresh the OST lock LVB from disk instead.
4937 Severity : enhancement
4939 Description: doing an ls at the same time as file IO can be slow
4940 Details : enqueue and other "small" requests can be blocked behind many
4941 large IO requests. Create a new OST IO portal for non-IO
4942 requests so they can be processed faster.
4945 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4947 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4948 created file with mode 000
4949 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4950 allow an ftruncate() as a non-root user to a file with mode 000.
4951 Lustre now respects this flag to disable mode checks when
4952 truncating a file owned by the user
4955 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4958 Description: Revoking locks from clients that went dead or catatonic might take
4960 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4961 cancellation of such locks instant on servers without waiting for
4962 any reply from clients. Clients drops these locks when cancel
4963 notification from server is received without replying.
4966 Frequency : liblustre-only, when liblustre client dies or becomes busy
4968 Description: Doing ls on Linux clients can take a long time with active
4970 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4971 granting such locks to it in the first place if possible. Locks
4972 are taken by proxy on the OST during the read or write and
4973 dropped immediately afterward. Add connect flags handling, do
4974 not grant locks to liblustre clients for glimpse ASTs.
4976 Severity : enhancement
4978 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4979 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4980 sequential reads and behaves badly with applications doing random
4981 reads. Improve it by reading ahead at least read region, and
4982 avoiding excessive large RPC for small reads.
4984 Severity : enhancement
4986 Description: Creating more than 1000 files for a single job may cause a load
4987 imbalance on the OSTs if there are also a large number of OSTs.
4988 Details : qos_prep_create() uses an OST index reseed value that is an
4989 even multiple of the number of available OSTs so that if the
4990 reseed happens in the middle of the object allocation it will
4991 still utilize the OSTs as uniformly as possible.
4996 Description: OST or MDS may oops in ping_evictor_main()
4997 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4998 but doesn't restart at beginning of obd_exports_timed list
5001 Severity : enhancement
5003 Description: improve by-nid export eviction on the MDS and OST
5004 Details : allow multiple exports with the same NID to be evicted at one
5005 time without re-searching the exports list.
5008 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5010 Description: MDS may oops in groups_free()
5011 Details : in rare race conditions a newly allocated group_info struct is
5012 freed again, and this can be NULL. The 2.4 compatibility code
5013 for groups_free() checked for a NULL pointer, but 2.6 did not.
5016 Frequency : common for liblustre clients doing little filesystem IO
5017 Bugzilla : 9352, 7313
5018 Description: server may evict liblustre clients accessing contended locks
5019 Details : if a client is granted a lock or receives a completion AST
5020 with a blocking AST already set it would not reply to the AST
5021 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5024 Frequency : lfs setstripe, only systems with more than 160 OSTs
5026 Description: unable to set striping with a starting offset beyond OST 160
5027 Details : llapi_create_file() incorrectly limited the starting stripe
5028 index to the maximum single-file stripe count.
5031 Frequency : LDAP users only
5033 Description: lconf did not handle in-kernel recovery with LDAP properly
5034 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5036 Severity : enhancement
5038 Description: bind OST threads to NUMA nodes to improve performance
5039 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5040 node and do their allocations there to localize memory access
5042 Severity : enhancement
5044 Description: llmount can determine client NID directly from Myrinet (GM)
5045 Details : the client NID code from gmnalnid was moved directly into
5046 llmount, removing the need to use this or specifying the
5047 client NID explicitly when mounting GM clients with zeroconf
5050 Frequency : if client is started with down MDS
5052 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5053 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5054 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5055 ptlrpc_import_delayed_req() because we didn't check if the
5056 request was interrupted, and we also didn't break out of the
5057 event loop if there was no timeout
5062 Description: data loss during non-page-aligned writes to a single file from
5063 both multiple nodes and multiple threads on one node at same time
5064 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5065 inconsistency led to false short-reads, that were cached and later
5066 used by ->prepare_write() to fill in partially written page,
5067 leading to data loss.
5070 Frequency : always, if lconf --abort_recovery used
5072 Description: lconf --abort_recovery fails with 'Operation not supported'
5073 Details : lconf was attempting to abort recovery on the MDT device and not
5076 Severity : enhancement
5078 Description: remove cleanup logs
5079 Details : replace lconf-generated cleanup logs with lustre internal
5080 cleanup routines. Eliminates the need for client-cleanup and
5083 Severity : enhancement
5085 Description: add support for EAs (user and system) on lustre filesystems
5086 Details : it is now possible to store extended attributes in the Lustre
5087 client filesystem, and with the user_xattr mount option it
5088 is possible to allow users to store EAs on their files also
5090 Severity : enhancement
5092 Description: Add possibility (config option) to show minimal available OST free
5094 Details : When compiled with --enable-mindf configure option, statfs(2)
5095 (and so, df) will return least minimal free space available from
5096 all OSTs as amount of free space on FS, instead of summary of
5097 free spaces of all OSTs.
5099 Severity : enhancement
5101 Description: do not expand extent locks acquired on OST-side
5102 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5103 by server: they are not cached anyway.
5106 Frequency : when mmap is used/binaries executed from Lustre
5108 Description: Unmmap pages before throwing them away from read cache.
5109 Details : llap_shrink cache now attempts to unmap pages before discarding
5110 them (if unmapping failed - do not discard). SLES9 kernel has
5111 extra checks that trigger if this unmapping is not done first.
5116 Description: lconf didn't resolve symlinks before checking to see whether a
5117 given mountpoint was already in use
5120 Frequency : when migrating failover services
5121 Bugzilla : 6395, 9514
5122 Description: When migrating a subset of services from a node (e.g. failback
5123 from a failover service node) the remaining services would
5124 time out and evict clients.
5125 Details : lconf --force (implied by --failover) sets the global obd_timeout
5126 to 5 seconds in order to quickly disconnect, but this caused
5127 other RPCs to time out too quickly. Do not change the global
5128 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5130 Severity : enhancement
5131 Frequency : if MDS is started with down OST
5132 Bugzilla : 9439,5706
5133 Description: Allow startup/shutdown of an MDS without depending on the
5134 availability of the OSTs.
5135 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5136 Add appropriate locking and lov-osc refcounts for safe
5137 cleaning. Add osc abort_inflight calls in case the
5138 synchronize never started.
5141 Frequency : occasional (Cray XT3 only)
5143 Description: root not authorized to access files in CRAY_PORTALS environment
5144 Details : The client process capabilities were not honoured on the MDS in
5145 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5146 been accessed by an authorized user then root was able to access
5147 the file on the local client also. The root user capabilities
5148 are now allowed on the MDS, as this environment has secure UID.
5151 Frequency : occasional
5153 Description: ldiskfs "too long searching" message happens too often
5154 Details : A debugging message (otherwise harmless) prints too often on
5155 the OST console. This has been reduced to only happen when
5156 there are fragmentation problems on the filesystem.
5161 Description: Division by zero in statfs when all OSCs are inactive
5162 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5163 lov_statfs divides by value returned by lov_get_stripecnt().
5167 Bugzilla : 9489, 3273
5168 Description: First write from each client to each OST was only 4kB in size,
5169 to initialize client writeback cache, which caused sub-optimal
5170 RPCs and poor layout on disk for the first written file.
5171 Details : Clients now request an initial cache grant at (re)connect time
5172 and so that they can start streaming writes to the cache right
5173 away and always do full-sized RPCs if there is enough data.
5174 If the OST is rebooted the client also re-establishes its grant
5175 so that client cached writes will be honoured under the grant.
5180 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5181 Details : Now I/O RPCs go to different portal number and (presumably) fast
5182 lock requests (and glimses) and other RPCs get their own service
5183 threads pool that should be able to service those RPCs
5186 Severity : enhancement
5188 Description: Ability to exchange lustre version between client and servers and
5189 issue warnings at client side if client is too old. Also for
5190 liblustre clients there is ability to refuse connection of too old
5192 Details : New 'version' field is added to connect data structure that is
5193 filled with version info. That info is later checked by server and
5197 Frequency : rare, liblustre only.
5198 Bugzilla : 9296, 9581
5199 Description: Two simultaneous writes from liblustre at offset within same page
5200 might proceed at the same time overwriting eachother with stale
5202 Details : I/O lock withing llu_file_prwv was released too early, before data
5203 actually was hitting the wire. Extended lock-holding time until
5204 server acknowledges receiving data.
5207 Frequency : extremely rare. Never observed in practice.
5209 Description: avoid generating lustre_handle cookie of 0.
5210 Details : class_handle_hash() generates handle cookies by incrementing
5211 global counter, and can hit 0 occasionaly (this is unlikely, but
5212 not impossible, because initial value of cookie counter is
5213 selected randonly). Value of 0 is used as a sentinel meaning
5214 "unassigned handle" --- avoid it. Also coalesce two critical
5215 sections in this function into one.
5217 Severity : enhancement
5219 Description: allow liblustre clients to delegate truncate locking to OST
5220 Details : To avoid overhead of locking, liblustre client instructs OST to
5221 take extent lock in ost_punch() on client's behalf. New connection
5222 flag is added to handle backward compatibility.
5224 Severity : enhancement
5225 Bugzilla : 4928, 7341, 9758
5226 Description: allow number of OST service threads to be specified
5227 Details : a module parameter allows the number of OST service threads
5228 to be specified via "options ost ost_num_threads={N}" in the
5229 OSS's /etc/modules.conf or /etc/modprobe.conf.
5233 Bugzilla : 6146, 9635, 9895
5234 Description: servers crash with bad pointer in target_handle_connect()
5235 Details : In rare cases when a client is reconnecting it was possible that
5236 the connection request was the last reference for that export.
5237 We would temporarily drop the export reference and get a new
5238 one, but this may have been the last reference and the export
5239 was just destroyed. Get new reference before dropping old one.
5241 Severity : enhancement
5242 Frequency : if client is started with failover MDS
5244 Description: Allow multiple MDS hostnames in the mount command
5245 Details : Try to read the configuration from all specified MDS
5246 hostnames during a client mount in case the "primary"
5249 Severity : enhancement
5251 Description: Stop sending data to evicted clients as soon as possible.
5252 Details : Check if the client we are about to send or are sending data to
5253 was evicted already. (Check is done every second of waiting,
5254 for which l_wait_event interface was extended to allow checking
5255 of exit condition at specified intervals).
5258 Frequency : rare, normally only when NFS exporting is done from client
5260 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5262 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5265 Frequency : always, if srand() is called before liblustre initialization
5267 Description: Liblustre uses system PRNG disturbing its usage by user application
5268 Details : Introduce internal to lustre fast and high-quality PRNG for
5269 lustre usage and make liblustre and some other places in generic
5270 lustre code to use it.
5272 Severity : enhancement
5273 Bugzilla : 9477, 9557, 9870
5274 Description: Verify that the MDS configuration logs are updated when xml is
5275 Details : Check if the .xml configuration logs are newer than the config
5276 logs stored on the MDS and report an error if this is the case.
5277 Request --write-conf, or allow starting with --old_conf.
5279 Severity : enhancement
5281 Description: Handle symlinks in the path when checking if Lustre is mounted.
5282 Details : Resolve intermediate symlinks when checking if a client has
5283 mounted a filesystem to avoid duplicate client mounts.
5288 Description: lconf can hit an error exception but still return success.
5289 Details : The lconf command catches the Command error exception at the top
5290 level script context and will exit with the associated exit
5291 status, but doesn't ensure that this exit status is non-zero.
5296 Description: failure of ptlrpc thread startup can cause oops
5297 Details : Starting a ptlrpc service thread can fail if there are a large
5298 number of threads or the server memory is very fragmented.
5299 Handle this without oopsing.
5302 Frequency : always, only if liblustre and non-default acceptor port was used
5304 Description: liblustre cannot connect to servers with non-default acceptor port
5305 Details : tcpnal_set_default_params() was not called and was therefore
5306 ignoring the environment varaible TCPNAL_PORT, as well as other
5307 TCPNAL_ environment variables
5312 Description: two objects could be created on the same OST for a single file
5313 Details : If an OST is down, in some cases it was possible to create two
5314 objects on a single OST for a single file. No problems other
5315 than potential performance impact and spurious error messages.
5319 Bugzilla : 5681, 9562
5320 Description: Client may oops in ll_unhash_aliases
5321 Details : Client dcache may become inconsistent in race condition.
5322 In some cases "getcwd" can fail if the current directory is
5328 Description: Inode refcounting problems in NFS export code
5329 Details : link_raw functions used to call d_instantiate without obtaining
5330 extra inode reference first.
5334 Bugzilla : 9942, 9903
5335 Description: Referencing freed requests leading to crash, memleaks with NFS.
5336 Details : We used to require that call to ll_revalidate_it was always
5337 followed by ll_lookup_it. Also with revalidate_special() it is
5338 possible to call ll_revalidate_it() twice for the same dentry
5339 even if first occurence returned success. This fix changes semantic
5340 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5341 reference on a request referred from the intent.
5342 ll_intent_release() then releases such a request.
5345 Frequency : rare, normally benchmark loads only
5347 Description: unlinked inodes were kept in memory on the client
5348 Details : If a client is repeatedly creating and unlinking files it
5349 can accumulate a lot of stale inodes in the inode slab cache.
5350 If there is no other client load running this can cause the
5351 client node to run out of memory. Instead flush old inodes
5352 from client cache that have the same inode number as a new inode.
5355 Frequency : SLES9 2.6.5 kernel and long filenames only
5356 Bugzilla : 9969, 10379
5357 Description: utime reports stale NFS file handle
5358 Details : SLES9 uses out-of-dentry names in some cases, which confused
5359 the lustre dentry revalidation. Change it to always use the
5363 Frequency : rare, unless heavy write-truncate concurrency is continuous
5364 Bugzilla : 4180, 6984, 7171, 9963, 9331
5365 Description: OST becomes very slow and/or deadlocked during object unlink
5366 Details : filter_destroy() was holding onto the parent directory lock
5367 while truncating+unlinking objects. For very large objects this
5368 may block other threads for a long time and slow overall OST
5369 responsiveness. It may also be possible to get a lock ordering
5370 deadlock in this case, or run out of journal credits because of
5371 the combined truncate+unlink. Solution is to do object truncate
5372 first in one transaction without parent lock, and then do the
5373 final unlink in a new transaction with the parent lock. This
5374 reduces the lock hold time dramatically.
5377 Frequency : rare, 2.4 kernels only
5379 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5380 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5381 before unmounting their filesystem in order to avoid deadlock.
5382 The kernel_locked() function in 2.4 kernels only checks whether
5383 the kernel lock is held, not whether it is this process that is
5384 holding it as 2.6 kernels do.
5389 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5390 Details : The client ptlrpc code may be trying to reconnect to a down
5391 server before a previous connection attempt has timed out.
5392 Increase the reconnect interval to be longer than the connection
5393 timeout interval to avoid sending duplicate connections to
5397 Frequency : echo_client brw_test command
5399 Description: fix echo_client to work with OST preallocated code
5400 Details : OST preallocation code (5137) didn't take echo_client IO path
5401 into account: echo_client calls filter methods outside of any
5402 OST thread and, hence, there is no per-thread preallocated
5403 pages and buffers to use. Solution: hijack pga pages for IO. As
5404 a byproduct, this avoids unnecessary data copying.
5408 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5409 Description: Client can oops in mdc_commit_close() after open replay
5410 Details : It was possible for the MDS to return an open request with no
5411 transaction number in mds_finish_transno() if the client was
5412 evicted, but without actually returning an error. Clients
5413 would later try to replay that open and may trip an assertion
5414 Simplify the client close codepath, and always return an error
5415 from the MDS in case the open is not successful.
5418 Frequency : rare, 2.6 OSTs only
5420 Description: OST may deadlock under high load on fragmented files
5421 Details : If there was a heavy load and highly-fragmented OST filesystems
5422 it was possible to have all the OST threads deadlock waiting on
5423 allocation of biovecs, because the biovecs were not released
5424 until the entire RPC IO was completed. Instead, release biovecs
5425 as soon as they are complete to ensure forward IO progress.
5427 Severity : enhancement
5429 Description: Support for specifying external journal device at mount
5430 Details : If an OST or MDS device is formatted with an external journal
5431 device, this device major/minor is stored in the ext3 superblock
5432 and may not be valid for failover. Allow detecting and
5433 specifying the external journal at mount time.
5438 Description: Mounting an MDS with pending unlinked files may cause oops
5439 Details : target_finish_recovery() calls mds_postrecov() which returned
5440 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5441 considers this an error and immediately begins cleaning up the
5442 lov, just after starting the mds_lov process
5444 Severity : enhancement
5446 Description: Implement 'lfs df' to report actual free space on per-OST basis
5447 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5448 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5449 usage of objects; '-h' to report in human readable format.
5451 ------------------------------------------------------------------------------
5453 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5460 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5461 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5462 at startup during orphan unlinking. Instead, skip buddy update
5463 before setup, as it will be regenerated after recovery is complete.
5468 Description: If an OST is inactive, its locks might reference stale inodes.
5469 Details : lov_change_cbdata() must iterate over all namespaces, even if
5470 they are inactive to clear inode references from the lock.
5472 Severity : enhancement
5473 Frequency : occasional, if non-standard max_dirty_mb used
5475 Description: Client will block write RPCs if not enough grant
5476 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5477 then the client will block writes while waiting for another RPC
5478 to complete instead of consuming its dirty limit. With change
5479 we get improved performance when max_dirty_mb is small.
5481 Severity : enhancement
5482 Bugzilla : 3389, 6253
5483 Description: Add support for supplementary groups on the MDS.
5484 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5485 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5486 lookups for user supplementary groups into a cache.
5490 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5491 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5492 for liblustre O_EXCL is enforced by MDS.
5496 Description: readdir never returns NULL in liblustre.
5497 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5498 dentry in struct dirent.
5502 Frequency : liblustre only, depends on application IO pattern
5503 Description: liblustre clients evicted if not contacting servers
5504 Details : Don't put liblustre clients into the ping_evictor list, so
5505 they will not be evicted by the pinger ever.
5507 Severity : enhancement
5509 Description: Add ability to evict clients by NID from MDS.
5510 Details : By echoing "nid:$NID" string into
5511 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5512 $NID would be instantly evicted from this MDS and from all active
5513 OSTs connected to it.
5517 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5518 Details : lookup_it_finish() used to query file size from OSTs that was not
5523 Description: service threads change working directory to that of init
5524 Details : Starting lustre service threads may pin the working directory
5525 of the parent thread, making that filesystem busy. Threads
5526 now change to the working directory of init to avoid this.
5530 Frequency : during shutdown only
5531 Description: shutdown with a failed MDS or OST can cause unmount to hang
5532 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5537 Frequency : 2.6 only
5538 Description: chmod/chown may include an extra supplementary group
5539 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5540 supplementary group and if none is specified this is used.
5543 Bugzilla : 5479 (6816)
5545 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5546 Details : If lookup is in progress on a file that is unlinked we might try
5547 to revalidate the inode and fail in revalidate after lookup is
5548 complete and ll_file_open() enqueues the open again but
5549 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5552 Frequency : always, if lconf --abort_recovery used
5554 Description: lconf --abort_recovery fails with 'Operation not supported'
5555 Details : lconf was attempting to abort recovery on the MDT device and not
5558 ------------------------------------------------------------------------------
5560 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5565 Frequency : rare (only unsupported configurations with a node running as an
5567 Bugzilla : 6514, 5137
5568 Description: Mounting a Lustre file system on a node running as an OST could
5570 Details : OSTs now preallocates memory needed to write out data at
5571 startup, instead of when needed, to avoid having to
5572 allocate memory in possibly low memory situations.
5573 Specifically, if the file system is mounted on on OST,
5574 memory pressure could force it to try to write out data,
5575 which it needed to allocate memory to do. Due to the low
5576 memory, it would be unable to do so and the node would
5577 become unresponsive.
5579 Severity : enhancement
5581 Description: Addition of lconf --service command line option
5582 Details : lconf now accepts a '--service <arg>' option, which is
5583 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5585 Severity : enhancement
5587 Description: Failover mode is now the default for OSTs.
5588 Details : By default, OSTs will now run in failover mode. To return to
5589 the old behaviour, add '--failout' to the lmc line for OSTs.
5591 Severity : enhancement
5593 Description: Health checks are now provided for MDS and OSTs
5594 Details : Additional detailed health check information on MSD and OSTs
5595 is now provided through the procfs health_check value.
5598 Frequency : occasional, depends on IO load
5600 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5601 after numerous create/delete cycles
5602 Details : The ext3 inode allocation policy would not allocate new inodes
5603 very well on the OSTs because there are no new directories
5604 being created. Instead we look for groups with free space if
5605 the parent directories are nearly full.
5610 Description: Network or server problems during mount may cause partially
5611 mounted clients instead of returning an error.
5612 Details : The config llog parsing code may overwrite the error return
5613 code during mount error handling, returning success instead
5619 Description: MDS can fail to allocate large reply buffers
5620 Details : After long uptimes the MDS can fail to allocate large reply
5621 buffers (e.g. zconf client mount config records) due to memory
5622 fragmentation or consumption by the buffer cache. Preallocate
5623 some large reply buffers so that these replies can be sent even
5624 under memory pressure.
5628 Frequency : rare (liblustre)
5629 Description: fsx running with liblustre complained that using truncate() to
5630 extend the file doesn't work. This patch corrects that issue.
5631 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5632 ATTR_SIZE and lsm use in llu_setattr_raw.
5636 Frequency : rare, only 2.6 kernels
5637 Description: Unusual file access patterns on the MDS may result in inode
5638 data being lost in very rare circumstances.
5639 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5640 code in the 2.6 kernel caused the inode and/or EA data not to
5641 be read from disk, causing single-file corruption.
5645 Frequency : rare, only 2.6 filesystems using extents
5646 Description: Heavy concurrent write and delete load may cause data corruption.
5647 Details : It was possible under high-load situations to have an extent
5648 metadata block in the block device cache from a just-unlinked
5649 file overwrite a newly-allocated data block. We now unmap any
5650 metadata buffers that alias just-allocated data blocks.
5654 Frequency : filesystems with default stripe_count larger than 77
5655 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5656 Details : lconf specifies an inode size of 4096 bytes when the default
5657 stripe_count is larger than 77. This conflicts with the default
5658 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5659 this case to avoid pinning too much memory for large EAs.
5661 ------------------------------------------------------------------------------
5663 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5668 Frequency : rare (extremely heavy IO load with hundreds of clients)
5670 Description: Client is evicted, gets IO error writing to file
5671 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5672 caused clients to be evicted for AST timeouts. The fixes in
5673 bug 5192 mean we no longer need to have such short AST timeouts
5674 so ldlm_timeout has been increased.
5677 Frequency : occasional during --force or --failover shutdown under load
5678 Bugzilla : 5949, 4834
5679 Description: Server oops/LBUG if stopped with --force or --failover under load
5680 Details : a collection of import/export refcount and cleanup ordering
5681 issues fixed for safer force cleanup
5684 Frequency : only filesystems larger than 120 OSTs
5685 Bugzilla : 5990, 6223
5686 Description: lfs getstripe would oops on a very large filesystem
5687 Details : lov_getconfig used kfree on vmalloc'd memory
5690 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5691 Bugzilla : 6242, 6243
5692 Description: reading from files that had been truncated to a non-zero size
5693 but never opened returned no data
5694 Details : ll_file_read() reads zeros from no-object files to EOF
5699 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5700 Details : There's little harm in aborting MDS/OSS recovery and letting it
5701 try again, so I removed the LASSERT and return an error instead.
5703 Severity : enhancement
5705 Description: New debugging infrastructure for tracking down data corruption
5706 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5707 (b) cover more of the client-side code path, and (c) try to narrow
5708 down where problems occurred
5712 Bugzilla : 3819, 4364, 4397, 6313
5713 Description: Racing close and eviction MDS could cause assertion in mds_close
5714 Details : It was possible to get multiple mfd references during close and
5715 client eviction, leading to one thread referencing a freed mfd.
5717 Severity: : enhancement
5718 Bugzilla : 3262, 6359
5719 Description: Attempts to reconnect to servers are now more aggressive.
5720 Details : This builds on the enhanced upcall-less recovery that was added
5721 in 1.4.2. When trying to reconnect to servers, clients will
5722 now try each server in the failover group every 10 seconds. By
5723 default, clients would previously try one server every 25 seconds.
5728 Description: After recovery, certain operations trigger a failed
5729 assertion on a client.
5730 Details : Failing over an mds, using lconf -d --failover, while a
5731 client was doing a readdir() call would cause the client to
5732 LBUG after recovery completed and the readdir() was resent.
5734 Severity : enhancement
5736 Description: Default groups are now added by lconf
5737 Details : You can now run lconf --group <servicename> without having to
5738 manually add groups with lmc.
5741 Frequency : occasional
5743 Description: Nodes with an elan id of 0 trigger a failed assertion
5746 Frequency : always when accessing e.g. tty/console device nodes
5748 Description: tty and some other devices nodes cannot be used on lustre
5749 Details : file's private_data field is used by device data and lustre
5750 values in there got lost. New field was added to struct file to
5751 store fs-specific private data.
5754 Frequency : when exporting Lustre via NFS
5756 Description: NFSD failed occasionally when looking up a path component
5757 Details : NFSD is looking up ".." which was broken in ext3 directories
5758 that had grown large enough to become hashed.
5761 Frequency : Clusters with multiple interfaces not on the same subnet
5763 Description: Nodes will repeatedly try to reconnect to an interface which it
5764 cannot reach and report an error to the log.
5765 Details : Extra peer list entries will be created by lconf with some peers
5766 unreachable. lconf now validates the peer before adding it.
5769 Frequency : Only if a default stripe is set on the filesystem root.
5771 Description: Setting a default stripe on the filesystem root prevented the
5772 filesystem from being remounted.
5773 Details : The client was sending extra request flags in the root getattr
5774 request and did not allocate a reply buffer for the dir EA.
5777 Frequency : occasional, higher if lots of files are accessed by one client
5778 Bugzilla : 6159, 6097
5779 Description: Client trips assertion regarding lsm mismatch/magic
5780 Details : While revalidating inodes the VFS looks up inodes with ifind()
5781 and in rare cases can find an inode that is being freed.
5782 The ll_test_inode() code will free the lsm during ifind()
5783 when it finds an existing inode and then the VFS later attaches
5784 this free lsm to a new inode.
5788 Bugzilla : 6422, 7030
5789 Description: MDS deadlock between mkdir and client eviction
5790 Details : Creating a new file via mkdir or mknod (starting a transaction
5791 and getting the ns lock) can deadlock with client eviction
5792 (gets ns lock and trying to finish a synchronous transaction).
5795 Frequency : occasional
5796 Description: While starting a server, the fsfilt_ext3 module could not be
5798 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5799 kernels. Previously, lconf would still use the ext3 name
5800 when trying to load modules. Now, it will correctly use
5801 ext3 on 2.4 and ldiskfs on 2.6.
5803 Severity : enhancement
5804 Description: The default stripe count has been changed to 1
5805 Details : The interpretation of the default stripe count (0, to lfs
5806 or lmc) has been changed to mean striping across a single
5807 OST, rather than all available. For general usage we have
5808 found a stripe count of 1 or 2 works best.
5810 Severity : enhancement
5811 Description: Add support for compiling against Cray portals.
5812 Details : Conditional compiling for some areas that are different
5816 Frequency : occasional
5817 Bugzilla : 6409, 6834
5818 Description: Creating files with an explicit stripe count may lead to
5819 a failed assertion on the MDS
5820 Details : If some OSTs are full or unavailable, creating files may
5821 trigger a failed assertion on the MDS. Now, Lustre will
5822 try to use other servers or return an error to the
5826 Frequency : occasional
5828 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5829 caused lock timeout during readahead (since 1.4.2).
5830 Details : Processes doing readahead might match a lock that hasn't been
5831 granted yet if there are overlapping and conflicting lock
5832 requests. The readahead process waits on ungranted lock
5833 (original lock is CBPENDING), while OST waits for that process
5834 to cancel CBPENDING read lock and eventually evicts client.
5836 Severity : enhancement
5838 Description: Initial enabling of flock support for clients
5839 Details : Implements fcntl advisory locking and file status functions.
5840 This feature is provided as an optional mount flag (default
5841 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5842 locking are implemented yet, and those that are are not guaranteed
5843 to be completely correct in production environments.
5844 mount -t lustre -o [flock|noflock] ...
5847 Frequency : occasional
5849 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5850 assertion in the ext3 JBD (journaling) layer.
5851 Details : The b_committed_data struct is protected by the big kernel lock
5852 in 2.4 kernels, serializing journal_commit_transaction() and
5853 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5854 access to this struct. In 2.6 kernels there is finer grained
5855 locking to improve SMP performance of the JBD layer.
5859 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5864 Description: merge and cleanup kernel patches.
5867 -----------------------------------------------------------------------------
5869 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5871 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5872 versions of Lustre. Please update all servers and clients to
5873 version 1.4.2 or later at the same time. You must also run
5874 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5875 to update the configuration logs.
5877 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5878 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5879 - swab llog records of type '0' so we get proper header size/idx (5861)
5880 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5881 - fix rename of one directory over another leaking an inode (5953)
5882 - avoid SetPageDirty on 2.6 (5981)
5883 - don't re-add just-being-destroyed locks to the waiting list (5653)
5884 - when creating new directories, inherit the parent's custom
5885 striping settings if present parent (3048)
5886 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5887 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5888 - don't hold client locks on temporary worklist from l_lru (5666)
5889 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5890 - automatically evict dead clients (5921)
5891 - Update file size properly in create+truncate+fstat case (6196)
5892 - Do not unhash mountpoint dentries, do not allow removal of
5894 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5895 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5896 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5897 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5898 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5899 - Correctly handle reads of files with no objects (6243)
5900 - lctl recover will also mark a device active if deactivate used (5933)
5902 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5903 - allow --write-conf on an MDS with different nettype than client (5619)
5904 - don't write config llogs to MDS for mounts not from that MDS (5617)
5905 - lconf should create multiple TCP connections from a client (5201)
5906 - init scripts are now turned off by default; run chkconfig --on
5907 lustre and chkconfig --on lustrefs to use them
5908 - upcalls are no longer needed for clients to recover to failover
5910 - add --abort-recovery option to lconf to abort recovery on device
5912 - add support for an arbitrary number of OSTs (3026)
5913 - Quota support protocol changes.
5914 - forward compatibility changes to wire structs (6007)
5915 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5916 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5917 - improved "lustre status" script
5918 - initialize blocksize for non-regular files (6062)
5919 - added --disable-server and --disable-client configure options (5782)
5920 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5921 - Vanilla 2.4.29 support
5922 - increase maximum number of obd devices to 520 (6242)
5923 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5924 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5925 SLES 9 2.6.5 kernels
5926 - limit stripes per file to 160 (the maximum EA size) (6093)
5928 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5931 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5932 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5933 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5934 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5935 - small ext3 extents cleanups and fixes (5733)
5936 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5937 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5938 - avoid needless client->OST connect, fix handle mismatch (5317)
5939 - fix DLM error path that led to out-of-sync client, long delays (5779)
5940 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5941 - fix several locking issues related to i_size (5492,5624,5654,5672)
5942 - don't move pending lock onto export if it is already evicted (5683)
5943 - fix kernel oops when creating .foo in unlinked directory (5548)
5944 - fix deadlock in obdfilter statistics vs. object create (5811)
5945 - use time_{before,after} to avoid timer jiffies wrap (5882)
5946 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5947 - Do not leak request if server was not able to process it (5154)
5948 - If mds_open unable to find parent dir, make that negative lookup(5154)
5949 - don't create new directories with extent-mapping (5909, 5936)
5951 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5952 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5953 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5954 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5955 - support for RHEL4 kernel on i686 (5773)
5956 - provide error messages when incompatible logs are encountered (5898)
5958 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5959 * version 1.4.0.10 (1.4.1 release candidate 1)
5961 - don't keep a lock reference when lock is not granted (4238)
5962 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5963 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5964 - drop import inflight refcount on signal_completed_replay error (5255)
5965 - unlock page after async write error during send (3677)
5966 - handle missing objects in filter_preprw_read properly (5265)
5967 - no transno return for symlink open, don't save no-trasno open (3440)
5968 - don't try to complete elan receive that already failed (4012)
5969 - free RPC server reply state on error (5406)
5970 - clean up thread from ptlrpc_start_thread() on error (5160)
5971 - readahead could read extra page into cache that wasn't ejected (5388)
5972 - prevent races in class_attach/setup/cleanup/detach (5260)
5973 - don't dereference de->d_inode after l_dput of de (5458)
5974 - use "int" for stripe value returned from lock_to_stripe (5544)
5975 - mballoc allocation and error-checking fixes in 2.6 (5504)
5976 - block device patches to fix I/O request sizes in 2.6 (5482)
5977 - look up hostnames for IB nals (5602)
5978 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5979 - don't start multiple acceptors for the same port (5277)
5980 - fix incorrect LASSERT in mds_getattr_name (5635)
5981 - export a proc file for general "ping" checking (5628)
5982 - fix "lfs check" to not block when the MDS is down (5628)
5984 - service request history (4965)
5985 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5986 - create an "evict_client" /proc entry on OSTs, like the MDS has
5987 - fix mount usage message, return errors per mount(8) (5168)
5988 - change grep [] to grep "[]" in tests so they work in more UMLs
5989 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5990 - remove ancient LOV_MAGIC_V0 EA support (5047)
5991 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5992 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5993 - allow readahead allocations to fail when low on memory (5383)
5994 - mmap locking landed again, after considerable improvement (2828)
5995 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5997 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6000 - send OST transaction number in read/write reply to free req (4966)
6001 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6002 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6003 - the watchdog thread now runs as interruptible (5246)
6004 - flock/lockf fixes (but it's still disabled, pending 5135)
6005 - don't use EXT3 constants in llite code (5094)
6006 - memory shortage at startup could cause assertion (5176)
6008 - reorganization of lov code
6009 - single portals codebase
6011 - add extents/mballoc support (5025)
6012 - direct I/O reads in the obdfilter (4048)
6013 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6015 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6018 - send OST transaction number in read/write reply to free req (4966)
6019 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6020 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6021 - the watchdog thread now runs as interruptible (5246)
6022 - handle missing objects in filter_preprw_read properly (5265)
6023 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6024 - drop import inflight refcount on signal_completed_replay error (5255)
6025 - unlock page after async write error during send (3677)
6026 - return original error code on reconstructed replies (3761)
6027 - no transno return for symlink open, don't save no-trasno open (3440)
6029 - add pid to ldlm debugging output (4922)
6030 - bump the watchdog timeouts -- we can't handle 30sec yet
6031 - extra debugging for orphan dentry/inode bug (5259)
6033 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6036 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6037 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6038 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6039 - fix NULL dereference in /proc/sys/portals/routes (4827)
6040 - allow failed mdc_close() operations to be interrupted (4561)
6041 - stop precreate on OST before MDS would time out on it (4778)
6042 - don't send partial-page writes before EOF from client (4410)
6043 - discard client grant for sub-page writes on large-page clients (4520)
6044 - don't free dentries not owned by NFS code, check generation (4806)
6045 - fix lsm leak if mds_create_objects() fails (4801)
6046 - limit debug_daemon file size, always print CERROR messages (4789)
6047 - use transno after validating reply (3892)
6048 - process timed out requests if import state changes (3754)
6049 - update mtime on OST during writes, return in glimpse (4829)
6050 - add mkfsoptions to LDAP (4679)
6051 - use ->max_readahead method instead of zapping global ra (5039)
6052 - don't interrupt __l_wait_event() during strace
6054 - add software watchdogs to catch hung threads quickly (4941)
6055 - make lustrefs init script start after nfs is mounted
6056 - fix CWARN/ERROR duplication (4930)
6057 - return async write errors to application if possible (2248)
6058 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6059 - print NAL number in %x format (4645)
6060 - update barely-supported suse-2.4.21-171 series (4842)
6061 - support for sles 9 %post scripts
6062 - support for building 2.6 kernel-source packages
6063 - support for sles km_* packages
6065 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6068 - ignore -ENOENT errors in osc_destroy (3639)
6069 - notify osc create thread that OSC is being cleaned up (4600)
6070 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6071 - reconstruct ost_handle() like mds_handle() (4657)
6072 - create a new thread to do import eviction to avoid deadlock (3969)
6073 - let lconf resolve symlinked-to devices (4629)
6074 - don't unlink "objects" from directory with default EA (4554)
6075 - hold socknal file ref over connect in case target is down (4394)
6076 - allow more than 32000 subdirectories in a single directory (3244)
6077 - fix blocks count for O_DIRECT writes (3751)
6078 - OST returns ENOSPC from object create when no space left (4539)
6079 - don't send truncate RPC if file size isn't changing (4410)
6080 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6081 - bind to privileged port in socknal and tcpnal (3689)
6083 - rate limit CERROR/CWARN console message to avoid overload (4519)
6084 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6085 - basic mmap support (3918)
6086 - kernel patch series update from b1_4 (4711)
6088 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6091 - avoid crash during MDS cleanup with OST shut down (2775)
6092 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6093 - don't use bad inodes on the MDS (3744)
6094 - dynamic object preallocation to improve recovery speed (4236)
6095 - don't hold spinlock over lock dumping or change debug flags (4401)
6096 - don't zero obd_dev when it is force cleaned (3651)
6097 - print grants to console if they go negative (4431)
6098 - "lctl deactivate" will stop automatic recovery attempts (3406)
6099 - look for existing locks in ldlm_handle_enqueue() (3764)
6100 - don't resolve lock handle twice in recovery avoiding race (4401)
6101 - revalidate should check working dir is a directory (4134)
6103 - don't always mark "slow" obdfilter messages as errors (4418)
6105 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6108 - don't close LustreDB during write_conf until it is done (3860)
6109 - fix typo in lconf for_each_profile (3821)
6110 - allow dumping logs from multiple threads at one time (3820)
6111 - don't allow multiple threads in OSC recovery (3812)
6112 - fix debug_size parameters (3864)
6113 - fix mds_postrecov to initialize import for llog ctxt (3121)
6114 - replace config semaphore with spinlock (3306)
6115 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6116 - don't allow enqueue to complete on a destroyed export (3822)
6117 - down write_lock before checking llog header bitmap (3825)
6118 - recover from lock replay timeout (3764)
6119 - up llog sem before sending rpc (3652)
6120 - reduce ns lock hold times when setting kms (3267)
6121 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6122 - fix NULL deref and obd_dev leak on setup error (3312)
6123 - replace some LBUG about llog ops with error handling (3841)
6124 - don't match INVALID dentries from d_lookup and spin (3784)
6125 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6126 - fix invalid assertion in ptlrpc_set_wait (3880)
6128 - add libwrap support for the TCP acceptor (3996)
6129 - add /proc/sys/portals/routes for non-root route listing (3994)
6130 - allow setting MDS UUID in .xml (2580)
6131 - print the stack of a process that LBUGs (4228)
6133 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6136 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6137 - make sure to unset replay flag from failed open requests (3440)
6138 - if default stripe count is 0, use OST count for inode size (3636)
6139 - update parent mtime/ctime on client for create/unlink (2611)
6140 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6141 - free recovery state on server during a forced cleanup (3571)
6142 - unregister_reply for resent reqs (3063)
6143 - loop back devices mounting and status check on 2.6 (3563)
6144 - fix resource-creation race that can provoke i_size == 0 (3513)
6145 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6146 - more debugging for page-accounting assertion (3746)
6147 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6148 - avoid deadlock after precreation failure (3758)
6149 - fix race and lock order deadlock in orphan handling (3450, 3750)
6150 - add validity checks when grabbing inodes from l_ast_data (3599)
6152 - add /proc/.../recovery_status to obdfilter (3428)
6153 - lightweight CDEBUG infrastructure, debug daemon (3668)
6154 - change default OSC RPC parameters to be better on small clusters
6155 - turn off OST read cache for files smaller than 32MB
6156 - install man pages and include them in rpms (3100)
6157 - add new init script for (un)mounting lustre filesystems (2593)
6158 - run chkconfig in %post for init scripts (3701)
6159 - drop scimac NAL (unmaintained)
6161 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6164 - clean kiobufs before and after use (3485)
6165 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6166 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6167 - change init script to fail more clearly if not run as root (1528)
6168 - allow clients to reconnect during replay (1742)
6169 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6170 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6171 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6172 - fix cond_resched (3554)
6173 - only evict unfinished clients after recovery (3515)
6174 - allow bulk resend, prevent data loss (3570)
6175 - dynamic ptlrpc request buffer allocation (2102)
6176 - don't allow unlinking open directory if it isn't empty (2904)
6177 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6178 - remove extraneous obd dereference causing LASSERT failure (3334)
6179 - don't use get_cycles() when creating temp. files on the mds (3156)
6180 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6181 - handle EEXIST for set-stripe, set proper directory name (3336)
6183 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6184 - fix ksocknal_fmb_callback() error messages (2918)
6186 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6189 - don't copy lvb into (possibly NULL) reply on error (2983)
6190 - don't deref dentry after dput, don't free lvb on error (2922)
6191 - use the kms to determine writeback rpc length (2947)
6192 - increment oti_logcookies when osc is inactive (2948)
6193 - update client's i_blocks count via lvb messages (2543)
6194 - handle intent open/close of special files properly (1557)
6195 - mount MDS with errors=remount-ro, like obdfilter (2009)
6196 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6197 - don't use cancelling-locks' kms values (2947)
6198 - use highest lock extent for kms, not last one (2925)
6199 - don't dereference ERR_PTR() dentry in error handling path (3107)
6200 - fix thread race in portals_debug_dumplog() (3122)
6201 - create lprocfs device entries at setup instead of at attach (1519)
6202 - common AST error handler, don't evict client on completion race (3145)
6203 - zero nameidata in detach_mnt in 2.6 (3118)
6204 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6205 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6206 - initialize RPC timeout timer earlier for 2.6 (3219)
6207 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6208 - print nal/nid for unknown nid (3258)
6209 - additional checks for oscc recovery before doing precreate (3284)
6210 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6211 - don't crash in mdc_close for bad permissions on open (3285)
6212 - zero i_rdev for non-device files (3147)
6213 - clear page->private before handing to FS, better assertion (3119)
6214 - tune the read pipeline (3236)
6215 - fix incorrect decref of invalidated dentry (2350)
6216 - provide read-ahead stats and refine rpc in flight stats (3328)
6217 - don't hold journal transaction open across create RPC (3313)
6218 - update atime on MDS at close time (3265)
6219 - close LDAP connection when recovering to avoid server load (3315)
6220 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6221 - don't leak open file on MDS after open resend (3325)
6222 - serialize filter_precreate and filter_destroy_precreated (3329)
6223 - loop device shouldn't call sync_dev() for nul device (3092)
6224 - clear page cache after eviction (2766)
6225 - resynchronize MDS->OST in background (2824)
6226 - refuse to mount the same filesystem twice on same mountpoint (3394)
6227 - allow llmount to create routes for mounting behind routers (3320)
6228 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6229 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6230 - fix rare problem with rename on htree directories (3417)
6232 - allow default OST striping configuration per directory (1414)
6233 - fix compilation for qswnal for 2.6 kernels (3125)
6234 - increase maximum number of MDS request buffers for large systems
6235 - change liblustreapi to be useful for external progs like lfsck (3098)
6236 - increase local configuration timeout for slow disks (3353)
6237 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6239 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6242 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6243 - don't overwrite extent policy data in reply if lock was blocked (2901)
6244 - drop filter export grants atomically with removal from device (2663)
6245 - del obd_self_export from work_list in class_disconnect_exports (2908)
6246 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6247 - swab reply message in mdc_close, other PPC fixes (2464)
6248 - fix destroying of named logs (2325)
6249 - overwrite old logs when running lconf --write_conf (2264)
6250 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6251 - fix race in target_handle_connect (2898)
6252 - mds_reint_create() should take same inode create lock (2926)
6253 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6254 - don't close files for self_export to avoid uninitialized obd (2936)
6255 - allow MDS with the same name as client node (2939)
6256 - hold dentry reference for closed log files for unlink (2325)
6257 - reserve space for all logs during transactions (2059)
6258 - don't evict page beyond end of stripe extent (2925)
6259 - don't oops on a deleted current working directory (2399)
6260 - handle hard links to targets without a parent properly (2517)
6261 - don't dereference NULL lock when racing during eviction (2867)
6262 - don't grow lock extents when lots of conflicting locks (2919)
6264 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6267 - account for cache space usage on clients to avoid data loss (974)
6268 - lfsck support in lustre kernel code (2349)
6269 - reduce journal credits needed for BRW writes (2370)
6270 - orphan handling to avoid losing space on client/server crashes
6271 - ptlrpcd can be blocked, stopping ALL progress (2477)
6272 - use lock value blocks to assist in proper KMS, faster stat (1021)
6273 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6274 - recovery for initial connections (2355)
6275 - fixes for mds_cleanup_orphans (1934)
6276 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6277 - block all file creations until orphan recovery completes (1901)
6278 - client remove rq_connection from request struct (2423)
6279 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6280 - recovery timer race (2670)
6281 - mdc_close recovey bug (2532)
6282 - ptlrpc cleanup bug (2710)
6283 - mds timeout on local locks (2588)
6284 - namespace lock held during RPCs (2431)
6285 - handle interrupted sync write properly (2503)
6286 - don't try to handle a message that hasn't been replied to (2699)
6287 - client assert failure during cleanup after abort recovery (2701)
6288 - leak mdc device after failed mount (2712)
6289 - ptlrpc_check_set allows timedout requests to complete (2714)
6290 - wait for inflight reqs when ptlrpcd finishes (2710)
6291 - make sure unregistered services are removed from the srv_list
6292 - reset bulk XID's when resending them (caught by 1138 test)
6293 - unregister_bulk after timeout
6294 - fix lconf error (2694)
6295 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6296 - readahead locks pages, leaves pending causing memory pressure (2673)
6297 - increase OST request buffers to 4096 on large machines (2729)
6298 - fix up permission of existing directories in simple_mkdir (2661)
6299 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6300 - don't assign transno to errored transactions (2742)
6301 - don't delete objects on OST if given a bogus objid from MDS (2751)
6302 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6303 - if rq_no_resend, then timeout request after recovery (2432)
6304 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6305 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6306 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6307 - MDS recovery shouldn't delete objects that it has given out (2730)
6308 - if enqueue arrives after completion, don't clobber LVB (2819)
6309 - don't unlock pages twice when trigger_group_io returns error (2814)
6310 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6311 - don't write pages to disk if there was an error (1450)
6312 - don't ping imports that have recovery disabled (2676)
6313 - take buffered bytes into account when balancing socknal conn (2817)
6314 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6315 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6316 - remove little-endian swabbing of llog records (1987)
6317 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6318 - retry reposting request buffers if they fail (1191)
6319 - grow extent at grant time to avoid granting a revoked lock (2809)
6320 - lock revoke doesn't evict page if covered by a second lock (2765)
6321 - disable VM readahead to avoid reading outside lock extents (2805)
6323 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6324 - updated kernel patches for hp-2.4.20 kernel (2681)
6326 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6329 - fix truncated write corruption (2366)
6330 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6332 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6333 - ensure len(uuid) < 37 in lmc (1171)
6334 - fix ia64 OOPS in llog_test (2255)
6335 - zero end of page at obdfilter for partial page writes (2648)
6336 - don't leave stale dentries around after renames (bug 2428)
6337 - fix timeouts when evicting a client with a single lock held (2642)
6338 - set deadline for the initial HELLO message to drain (2634)
6339 - print out dotted-quad IP addresses in the socknal (2302)
6341 - additional debugging for MDS client eviction problem (2443)
6342 - fix mkfsoptions support for osts (2603, 2604)
6344 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6347 - add series for the vanilla 2.6.0 kernel
6348 - add series for the vanilla 2.4.24 kernel
6349 - add series for a cray x86/64 UL kernel drop
6350 - fix xattr patches for the vanilla 2.4.19 series
6352 - generate true UUIDs in lmc (1171)
6353 - have portals stack dumping break in UML (2466)
6354 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6355 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6356 - wake callback waiting threads on client eviction (2460)
6357 - Add --ptldebug and --subsystem to lmc (1719)
6358 - update assertion to allow safe interrupt allocation
6359 - set rq_no_resend for cancel requests (2432)
6360 - recalculate ptlrpcd timeout after resend (2494)
6361 - call vfs_rmdir when removing pending directories (2368)
6362 - fix renaming a file to itself (2429)
6363 - lmc creates a default one-stripe lov (2454)
6364 - expand procfs space to handle large clusters (2326)
6365 - increase UML stack to avoid overflow
6366 - update lconf's list of debug and subsystem masks
6367 - fix lfs find --obd (2510)
6368 - /proc tunable for disabling filter read caching (2591)
6369 - stop rpm packages from altering slapd.conf (2301)
6370 - disable nagle in the socknal under 0conf (2578)
6371 - choose mds inode size based on stripe count (2572)
6372 - fix kernel-source rpm problems (2516)
6374 - add --disable-doc to avoid pdf generation (2421)
6375 - update documentation, tests, type-os, comments
6376 - avoid format warnings on ia64
6377 - remove the TOE NAL
6378 - tiny code cleanups by removing unused fields
6380 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6383 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6384 - remove the most blatant lies from BUILDING file (2371)
6385 - change default debug level to reasonable production setting
6386 - reduce client side cache size to reduce cache flush time
6387 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6388 - make TCP zerocopy and pinger support enabled by default (2476)
6389 - sync writes completed after process exits caused crashes (2319)
6390 - maintain correct mount count on the MDS (2356)
6391 - backout 1557, because 2316 wasn't really fixed
6392 - better file I/O statistics gathering in /proc
6393 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6394 - another kernel patch to fix zero-copy TCP function export
6395 - don't take duplicate lock when processing re-sent getattr (2420)
6396 - lctl uses obd_self_export instead of creating new conn (2353)
6397 - MDS/OST recovery case which requires object creation asserted (2425)
6398 - move lfs from /usr/sbin to /usr/bin in packages
6399 - fix race between mds_client_add and mds_client_free (2417)
6400 - use kmalloc instead of slabs in portals (2430)
6401 - don't create duplicate records when a failover MDS is present (2442)
6402 - remove unnecessary mount age check (2332)
6403 - don't remove directory inodes from locks prematurely (2451)
6404 - don't break if MDS service name is the same as hostname (2103)
6405 - fix races in client write RPC generation when cache full (2482)
6407 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6410 - remove now-unused request->rq_obd (278)
6411 - if an allocation fails, print out how much memory we've used (1933)
6412 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6413 - add the "configurable stack size" patch to most series files (1256)
6414 - ability to write large log records, for 100+ OST configs (2306)
6415 - fix NULL deref when filter_prep fails (2314)
6416 - fix operator precedence error in filter_sync
6417 - dynamic allocation of socknal TX descriptors (2315)
6418 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6419 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6420 - ensure that CWARN messages in Portals always get to the syslog
6421 - __init/__exit are not for prototype decls (ldlm_init/exit)
6422 - x86-64 compile warning fixes
6423 - fix gateway LMC keyword conflict (2318)
6424 - fix MDS lock inversions in getattr/reint paths (1844)
6425 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6426 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6427 - DLM race condition prevented some lock evictions (2328)
6428 - ENOMEM detection and retry on socknal sends (2230)
6429 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6430 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6431 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6432 - better error messages when a client is rejected during recovery (1505)
6433 - avoid cancelling locks which were never granted, after failure (2330)
6434 - fix i_sem/journal inversion in mds_client_add (2333)
6435 - fix truncate/getattr lock cycle deadlock (2334)
6436 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6437 - fix two rare exit paths which could leak an l_lock() ref (2321)
6438 - fencepost error in MDS/OST orphan recovery (2226)
6439 - make log record alignment 8 bytes (1988)
6440 - lstripe now fails when requested offset > ost_count (2237)
6441 - ensure that all kernel series have a complete list.h (1607)
6442 - fix crashes in special-file operations (2316)
6443 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6445 - allow configurable automake binary, for testing new versions
6446 - small update to the lfs documentation
6448 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6450 * fix negative export reference count in fsfilt_sync (2312)
6452 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6453 * release candidate 0.9.1
6455 - orphans are moved into the PENDING directory for possible recovery
6456 - replayed opens now open by fid for orphan/rename safety (1042)
6457 - last close of an orphan inode generates a transno (683)
6458 - chdir() and mount() now pin the directory entry (1020)
6459 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6460 - discard very old requests without processing them (1502)
6461 - remove some common, well-understood CERRORs (1505)
6462 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6463 - clear "grant" flags in OST replies until OST grant code lands (1644)
6464 - fix read performance by not clobbering i_blksize on client (1598)
6465 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6466 - make LRU size a /proc tunable, clears locks when reduced (707)
6467 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6468 - server locks take references on exports now (1558)
6469 - build fixes for 2.4.20-rh trees (1663)
6470 - return an error from lov_create if all OSCs are inactive (1751)
6471 - fix import levels when a reconnect happens without a timeout (1597)
6472 - exit early from mds_open if we get a lookup error (1749)
6473 - partial page read at EOF wouldn't wait for disk before sending (1642)
6474 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6475 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6476 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6477 - write/truncate lock inversion (1639)
6478 - Don't auto-load obdclass, portals modules during cleanup (1495)
6479 - fix timestamps from jumping to "now" (1763)
6480 - extra journal assertions (1648)
6481 - add an extra multiunlink test (1771)
6482 - fix read_record/write_record API (1776)
6483 - fix leak of offset_extent, possible incorrect i_size later (1772)
6484 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6485 - Debugging for the kqswnal_get_idle_tx problems (1820)
6486 - Allow recovery to be attempted multiple times (1536)
6487 - Write out MDS last_rcvd file after it is first created (1600)
6488 - Fix tx_descriptor leak in failed transmit situations (1827)
6489 - ext3 journaling fixes for assertion failure after IO error (1871)
6490 - class_export_put() on freed export after completion AST error (1896)
6491 - Fix revalidate looping in VFS (1322)
6492 - Don't access a freed export during MDS_REINT timeout (1521)
6493 - Add open-unlink recovery support on the MDS (1673,1764)
6494 - Return an error if no MDS data was read from last_rcvd (1946)
6495 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6496 - Don't setup a disk device that doesn't match exported UUID (317)
6497 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6498 - avoid committing NULL handle in force close
6499 - local.sh is now a one-stripe LOV configuration
6500 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6501 - fix ext3 htree duplicate directory entry corruption (1516)
6502 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6503 - update to new LOV EA format (2097)
6504 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6506 2003-06-15 Phil Schwan <phil@clusterfs.com>
6509 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6510 - per-import/export recovery handling (958, 931, 959)
6511 - multiple last-rcvd slots, for serving multiple FSes (949)
6512 - connections are again shared between multiple imp/exports (963, 964)
6513 - "umount -f" would hang if any requests needed to be sent (393, 978)
6514 - avoid pinning large req buffer by copying for queued messages (989)
6515 - add "uuid" to "lctl device" command to help upcalls (991)
6516 - "open" RPCs with transnos would confuse recovery counters (1037)
6517 - do proper endian conversion of all wire messages (288, 340, 891)
6518 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6519 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6520 - fix ptlrpc_connection leak in target_handle_connect (1174)
6521 - fix import refcounting bug in OST and MDS cleanup (1134)
6522 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6523 - fix very unlikely obd_types race condition (501)
6524 - remove osc_open hack for echo_client (1187)
6525 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6526 - a failure in read_inode2 leads to deadlock (1139)
6527 - cancel ack-locks as soon as transaction is committed (1072)
6528 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6529 - make sure to commitrw after any preprw to avoid deadlock (1162)
6530 - failing to execute a file in a lustre FS would lock inode (1203)
6531 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6532 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6533 - obd_uuid2tgt would walk off the end of the list (1255)
6534 - on IA64 the osc would give portals incorrect bulk size (1258)
6535 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6536 - fix lock inversion caused by new llite matching code (1282)
6537 - limit the number of dirty pages on a client to 10MB (1286)
6538 - timed out locks were not being corrected cancelled (1289)
6539 - fix O_DIRECT above 4GB on IA-32 (1292)
6540 * major user-visible changes
6541 - fail out/fail over policy now controlled by the upcall (993)
6543 - add OBD_PING to check server availability and failure (954)
6544 - lustre messages are now sent in sending host order (288, 340, 891)
6545 - add eadatalen to MDS getattr reply (340)
6546 - OST read replies may contain second buffer, with per-page status (593)
6548 2003-03-11 Phil Schwan <phil@clusterfs.com>
6551 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6552 - failed open()s could cause deadlock; fixed (867, 869)
6553 - stop cancelling OST locks when files are closed (481)
6554 - overlapping XID spaces caused network corruption (851, 853)
6555 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6556 - setattr_raw added, to do single-RPC, server-side setattrs
6557 - lmc/lconf syntax change for OST UUIDs
6558 - fix crashy race condition between ptlrpc_free_req and osc_close
6559 - don't use request in mdc_enqueue if we hit a timeout (889)
6560 - don't set the inode i_size for regular files from the MDS (896)
6561 - handle out of order completion AST (842)
6562 - don't LBUG if a lock request times out after receiving AST (913)
6563 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6564 - if a bad lock AST arrives, send an error instead of dropping entirely
6565 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6566 - fix leak in bulk IO when only partially completed (899, 900, 926)
6567 - fix O_DIRECT for ia64 (55)
6568 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6569 - C-z after timeout could hang a process forever; fixed (977)
6571 - client-side I/O cache (678, 924, 929, 941, 970)
6573 - READPAGE and SETATTRs which don't take server-side locks get
6576 2003-02-11 Phil Schwan <phil@clusterfs.com>
6579 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6580 - Fix fsfilt_extN_readpage() to read a full page of directory
6581 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6582 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6583 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6584 - Most intent operations were removed, in favour of a new RPC mode
6585 that does a single RPC to the server and bypasses most of the VFS
6586 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6587 - Aggressively cancel local locks on DLM servers
6588 - mds_reint_unlink sends EA to the client if it's the last nlink.
6589 client uses that EA to unlink OST objects.
6590 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6591 - recursive symlinks were fixed (439)
6592 - fixed NULL deref in DEBUG_REQ
6593 - filter_update_lastobjid no longer calls sync, which annoyed extN
6594 - fixed multi-client small-writes to a single file problem (445)
6595 - fixed mtime updates during file writes (607)
6596 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6597 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6598 - fixed Linux OST size reporting problem (444, 656)
6599 - OST now updates object mtime with writes or setattr (607, 619)
6600 - client verifies file size before zeroing page past EOF (445)
6601 - OST now writes last allocated objid to disk with allocation (108)
6602 - LOV on echo now works (409)
6604 - mds_reint_unlink sends a new buffer, with the EA included. this
6605 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6606 set if a regular file was being unlinked, and it was the last link
6607 - use PtlGet from the target for bulk writes (315)
6608 - OST now updates object mtime with writes or setattr (607, 619)
6609 - LDLM now has a grant-time callback to revalidate locked items, if
6611 - Many MDS operations were reorganized to combat race conditions
6613 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6614 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6616 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6619 - Fully reactivate OST imports after reconnection (512, others)
6620 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6621 - More graceful error handling for truncating on dead OST (515)
6622 - Don't error out unless we're actually accessing dead stripes (474)
6623 - Fix garbage sizes when stripes are missing (410)
6624 - LRU counters were broken, causing constant lock purge (433, 432)
6625 - garbage on read from stripes with failed OSTs (441)
6626 - mark OSCs as active before reconnecting during recovery (438)
6627 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6628 - lfind did not preserve OST order in output (443)
6629 - symlinks cause hung clients, incorrect data (439)
6630 - stop dereferencing request after dropping refcount (457)
6631 - don't LASSERT(spin_is_locked) on non-SMP (455)
6632 - fixes for many rename() bugs
6633 - fstat didn't correctly synchronize attributes (399)
6634 - server must handle lock cancellation during blocking AST prep (487)
6635 - bulk descriptors were free()d too soon (511)
6636 - fix paths in lconf, which would load incorrect modules (451, 507)
6637 - fix confusing lconf 'host not found' error message (386)
6638 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6639 - fix race condition in mdc_blocking_ast() for inode access (526)
6640 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6641 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6642 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6643 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6644 - fix lstripe in several ways (kernel side) (527)
6645 - fix request leak in ldlm_cli_enqueue (262)
6646 - incorrect OSC was marked inactive after OST failure
6647 - call mds_fs_cleanup before unmounting filesystem (524)
6648 - fix races between taking ns_lock and ldlm_lock_change_resource
6649 - fix races updating LOV export open file list
6650 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6651 - fix recovery NULL deref in ldlm_cli_cancel_unused
6652 - fix some DLM races by using new hash table for lock handles (419)
6653 - permit the client to specify desired inodes, at replay
6654 - duplicate requests when we queue them for replay reintegration
6655 - fix last_rcvd offset calculation
6656 - sync after each recovered transaction, so we always make progress
6657 - never, not always, ERESTART requests without transnos
6658 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6660 - skip replay if the MDS says that the client is already connected
6661 - don't check for a recovery-enabled export to match lctl's UUID
6662 - don't INC_USE_COUNT for phantom exports
6663 - don't crash when cleaning up phantom exports (567)
6664 - don't double-finish or set replay data for errored mdc_open requests
6665 - abort requests when they time out, so we don't get old replies
6666 - send/receive replies for AST messages again
6667 - if the client says that it doesn't have the lock, cancel it on the
6669 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6670 mark it as destroyed, it will all work out when decref is called
6671 - fix module use counts (22, 581)
6673 - ASTs now expect a reply (server cancels lock on error reply)
6675 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6678 - fix many simultaneous client startup (392)
6679 - fix dentry->d_it clobbering
6680 - credentials weren't being shipped for readdir/getattr operations
6681 - remove invalid assertions triggered during some concurrent MD
6683 - proper Lustre versions added (336, 389)
6684 - fix memory leak for create error case (398)
6685 - fix LOV locking bug that would get cli/srv out of sync
6686 - fix echo client over LOV (409)
6687 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6688 - fix double-O_EXCL intent crash (424)
6689 - avoid sending multiple lock CANCELs (352)
6691 - MDS can do multi-client recovery (modulo bugs in new code)
6693 - many updates, edits, cleanups
6695 2002-11-18 Phil Schwan <phil@clusterfs.com>
6698 - fix null d_it dereference (346)
6699 - fix full OST/dbench hang (333)
6700 - fix permission problem with file removal (286)
6701 - fix removal of OSCs from LOV when they fail
6702 - fix NULL deref during bulk timeout (214)
6703 - fix problems related to multiple filesystems on one MDS (241)
6704 - fixed serious subtle metadata locking bugs
6705 - free locks on clients when inodes are removed due to memory
6707 - fix inode pointer in lock data (285)
6708 - partial support for multiple MDS on a single host (241)
6709 - data locks weren't cancelled at clear_inode time (290, 311)
6710 - intent locks could lead to unbounded lock growth (205)
6711 - added a maximum lock count, an LRU list, and a flusher
6712 - fix multiple rename (365)
6713 - properly abstracted the echo client
6714 - OSC locked 1 byte too many; fixed
6715 - rewrote brw callback code:
6716 - fixed recovery bugs related to LOVs (306)
6717 - fixed too-many-pages-in-one-write crash (191)
6718 - fixed (again) crash in sync_io_timeout (214)
6719 - probably fixed callback-related race (385)
6721 - Add capability to MDS protocol
6722 - LDLM cancellations and callbacks on different portals
6724 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6727 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6728 - defer open object destruction to close time (601981, 138)
6729 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6730 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6731 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6732 - don't call obd_setattr in ll_file_release for destroyed objects
6733 * protocol change to lustre_msg: move |version| and add |flags|
6734 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6735 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6736 * added replay of create, unlink, link and rename operations during
6737 MDS failover; recovery should be much more robust now
6738 * remove failed OSCs from LOVs (only lov_create uses this so far)
6739 * the lustre-HOWTO was brought (more) up to date (582544)
6741 2002-10-23 Phil Schwan <phil@clusterfs.com>
6744 - in-use dentries weren't being reused properly (617851)
6745 - prevent multiple LDLM setup (599178)
6746 - fix LOV size calculations for truncate (617853)
6747 - fix client handling of MDS intent errors (POSIX)
6748 - fix permission bug in lovstripe.c test (624321)
6749 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6750 - truncate past end of file could corrupt data
6751 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6752 - a race in recovery could return ETIMEDOUT to apps (623947)
6753 - building outside the source directory was fixed
6754 * the lustre-HOWTO was brought (more) up to date (582544)
6755 * major progress was made on recovery functionality
6757 2002-10-10 Phil Schwan <phil@clusterfs.com>
6760 - recovery deadlock fix
6761 - rm -rf causes LBUG fix (617817)
6762 - file open by multiple tasks fix (618962)
6763 - directory permissions bugs (602707 and 620007)
6764 - journal_stop fixed with locking (611313)
6765 - O_APPEND failures resolved (618273, perhaps 614459)
6766 - lconf PATH fix (619770)
6767 - IA64 build fix (621450)
6768 - RPC buffer sizes scale with amount of memory
6770 2002-10-01 Phil Schwan <phil@clusterfs.com>
6773 - locks would be cancelled without throwing away data pages,
6774 resulting in inconsistent data (605627)
6775 - inode attributes were not always being refreshed (605627, 612449)
6776 - lconf now continues to cleanup after lctl reports an error
6777 - MDS now enforces user permissions (602707)
6778 - lprocfs cleanup fixed, but not yet enabled (614157)
6779 - fixed infinite server hang, should a client not respond to an AST
6780 - avoid going into recovery if user calls readlink() with a buffer
6781 that's too small (613941)
6782 - AST RPCs no longer require replies (614867) -- this may be changed
6783 - don't crash server if client sends an IOV that's too big (611336)
6784 - fixed lock conversion deadlock (611892)
6785 - fixed the following of symlinks (614622)
6786 * recovery: the server can remove locks from a client that dies, other
6787 clients can make progress
6788 * more extN patch fixes
6789 * compile-time configurable ptlrpc buffer allocations
6791 - collaborative read cache document
6792 - Lustre Lite Performance CDR document-in-progress
6794 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6797 - fix typo in patch-2.4.18
6799 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6802 - clear ptlrpc request each time in handle_incoming_request()
6803 - unlink of files now destroys the object on the OST
6805 2002-09-19 Peter Braam <braam@clusterfs.com>
6807 * add hard link support
6808 * change obdfile creation method
6809 * kernel patch changed
6811 2002-09-19 Peter Braam <braam@clusterfs.com>
6814 - stack overflow bug in extN fixed
6816 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6818 * documentation updates
6819 - add man pages for config tools
6820 - update tests/README to describe testing with new config tools
6821 - finish metadata API descriptions
6822 * bug fixes and cleanups
6823 - statfs workaround for 16TB limit
6824 - LOV stripe allocation improved, can stripe on subset of OSTs
6825 - LOV file size/IO offset was wrong for files > 4GB in size
6826 - object EA data was being dropped, caused files to be unreadable
6827 - memory overflow with non-LOV OST caused memory corruption
6828 - fixed regression tests to work with new config tools, obdfilter
6829 - fixed bug when directory size became larger than 1 block
6830 - fixed bug (for single client case) when PWD was deleted
6831 - invalidate local directory pages when doing intent-based ops
6832 - avoid LDLM oops when lock callback contained bad data
6834 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6836 * documentation updates
6837 * bug fixes and cleanups
6838 - configuration tools
6841 - 64-bit compile warnings
6842 - 64-bit internal statfs data
6844 * test_brw on persistent OST devices
6846 * lprocfs (disabled)
6848 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6850 * documentation updates
6851 * bug fixes and cleanups
6852 * configuration tools
6854 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6857 * many small fixes to 0_5_4
6858 * io/network handling
6859 * thinkos in MDS operations
6861 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6864 * crucial basic fixes to 0.5.3
6865 * IOR, Iozone work over Elan
6868 2002-08-07 Phil Schwan <phil@clusterfs.com>
6869 * version 0_5_3, our first alpha
6870 * we use the new Portals iovs
6871 * documentation updates
6872 * bug fixes and cleanups
6873 * small changes in the DLM wire protocol
6875 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6876 * version 0_5_1 with some initial stability,
6877 * locking on MD and file I/O.
6878 * documentation updates
6879 * several bug fixes since 0.5.0
6880 * small changes in wire protocol
6882 2002-07-18 Phil Schwan <phil@clusterfs.com>
6884 * delivered as Lustre Light Alpha
6885 * fixed a crash after handling invalid MDS requests
6886 * fixed directory pages for architectures with non-4k pages sizes
6888 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6889 * release version v0_4_4
6890 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6892 * Fixes a number of interruption problems with OST operations.
6893 * Update documentation for portals header changes
6894 * Move all wire protocol structs/defines to lustre_idl.h
6895 * Fixes symlink length bug.
6896 * Add tcpdump to repository.
6898 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6899 * release version v0_4_3
6900 * Fixes statfs for inodes on extN.
6901 * Fixes bug in runtests which would delete /etc/hosts.
6902 * Use 64-bit object IDs wherever possible (not into VFS though)
6903 Remove ost_get_info, which is unused by lustre, and out of date.
6905 2002-07-03 Peter Braam <braam@clusterfs.com>
6906 * release version v0_4_2 Fixes a lookup error (type not passed)
6907 * move forward to head of Portals
6908 * move forward to latest Lustre kernel
6910 2002-06-25 Peter Braam <braam@clusterfs.com>
6911 * release version v0_4_1. Hopefully stable on single node use.